The Cambridge Companion to Chomsky

The Cambridge Companion to Chomsky

Noam Chomsky is one of the most influential thinkers of modern times. Themost cited writer in the humanities, his work has revolutionized the field oflinguistics, and has dominated many other disciplines including politics andthe philosophy of mind and human nature. He has also contributed significantlyto our understanding of the abuse of power, and of the controlling effects ofthe mass media.

This companion brings together a team of leading linguists, philosophers,cognitive scientists, and political theorists to consolidate the disparate strandsof Chomsky’s thought into one accessible volume. Through a range of chap-ters focusing on the various aspects of his work, they introduce in a clear andnon-technical way the central themes of his extraordinary effect on our under-standing of language, mind, and the abuse of political power, and provide anengaging insight into the connections between Chomsky’s work in each ofthese areas.

Comprehensive and informative, this is an essential guide to one of theleading intellectual figures of our time.

james mcgilvray is Associate Professor of Philosophy at McGill Uni-versity. His previous publications include Chomsky (1999), Tense, Reference,and Worldmaking (1991), and Social and Political Philosophy (co-edited withCharles King, 1973). He has also written for a variety of journals includingSynthese, Nous, Canadian Journal of Philosophy, Mind and Language, andPhilosophical Studies.

Cambridge Collections Online © Cambridge University Press, 2007


The Cambridge Companionto Chomsky

Edited by

James McGilvrayMcGill University


published by the press syndicate of the university of cambridgeThe Pitt Building, Trumpington Street, Cambridge, United Kingdom

cambridge university pressThe Edinburgh Building, Cambridge, CB2 2RU, UK40 West 20th Street, New York, NY 10011–4211, USA477 Williamstown Road, Port Melbourne, VIC 3207, AustraliaRuiz de Alarcon 13, 28014 Madrid, SpainDock House, The Waterfront, Cape Town 8001, South Africa

http://www.cambridge.org

C© Cambridge University Press 2005

This book is in copyright. Subject to statutory exceptionand to the provisions of relevant collective licensing agreements,no reproduction of any part may take place withoutthe written permission of Cambridge University Press.

Printed in the United Kingdom at the University Press, Cambridge

Typeface Times 10/12 pt. System LATEX 2ε [tb]

A catalogue record for this book is available from the British Library

Library of Congress Cataloguing in Publication dataThe Cambridge companion to Chomsky / edited by James McGilvray.

p. cm.ISBN 0 521 78013 6 (hardback) – ISBN 0 521 78431 X (paperback)1. Chomsky, Noam. I. McGilvray, James A. (James Alasdair), 1942–P85.C47C36 2005410′.92 – dc22 2004051104

ISBN 0 521 78013 6 hardbackISBN 0 521 78431 X paperback

The publisher has used its best endeavors to ensure that the URLs for externalwebsites are correct and active at the time of going to press. However, the publisherhas no responsibility for the websites and can make no guarantee that a site willremain live or that the content is or will remain appropriate.

First published 2005Reprinted 2005


Contents

The contributors page vii

Introduction 1james mcgilvray

Part I Chomsky on the human language

1 Chomsky’s science of language 21neil smith

2 Plato’s Problem, UG, and the language organ 42david lightfoot

3 Grammar, levels, and biology 60howard lasnik

4 How the brain begets language 84laura-ann petitto

5 Chomsky and Halle’s revolution in phonology 102b. elan dresher

6 Universal aspects of word learning 123lila gleitman and cynthia fisher

Part II Chomsky on the human mind

7 Empiricism and rationalism as research strategies 145norbert hornstein

8 Innate ideas 164paul pietroski and stephen crain

9 Mind, language, and the limits of inquiry 181akeel bilgrami and carol rovane

v


vi Contents

10 Meaning and creativity 204james mcgilvray

Part III Chomsky on values and politics

11 Market values and libertarian socialist values 225milan rai

12 The individual, the state, and the corporation 240james wilson

13 Noam Chomsky: the struggle continues 260irene gendzier

14 The responsibility of the intellectual 280jean bricmont

Notes 295References 313Index 331


The contributors

akeel bilgrami is Johnsonian Professor of Philosophy at ColumbiaUniversity. He specializes in the philosophies of mind and language. HisSelf-Knowledge and Intentionality and The Moral Psychology of Identity areforthcoming.

jean bricmont is Professor of Theoretical Physics at the Catholic Univer-sity of Louvain. He co-authored with Alan Sokal the volume FashionableNonsense. He often writes and speaks on political topics.

stephen crain is Professor of Linguistics at the University of Maryland,College Park. He is primarily interested in issues of child languagedevelopment – especially, recently, in children’s acquisition of semanticknowledge. He is co-author with Rozz Thornton of Investigations in Uni-versal Grammar.

b. elan dresher is Professor of Linguistics at the University of Toronto.He has written many articles on phonology. Among his interests is the historyof phonological theory.

cynthia fisher is an Associate Professor in the Department of Psychologyand Beckman Institute, University of Illinois at Urbana-Champaign. Herresearch focuses on child language acquisition, both words and linguisticstructures.

irene gendzier is Professor of Political Science at Boston University. Sheworks on issues of political economy and international political development,comparative politics, and the politics of the Middle East. She is the authorof Notes From the Minefield: United States Intervention in Lebanon and theMiddle East, 1945–1958 and Development Against Democracy.

lila gleitman is Steven and Marcia Roth Professor in the Department ofPsychology and Professor in the Department of Linguistics at the Univer-sity of Pennsylvania. She focuses on linguistic structure (morphological andsyntactic) and the acquisition of language (sounds, structures, and words).She is the editor of Invitation to Cognitive Science, vol. I, Language and,

vii


viii The contributors

with Barbara Landau, author of Language and Experience: Evidence Fromthe Blind Child.

norbert hornstein is Professor of Linguistics at the University of Mary-land. His research currently focuses on issues of Chomsky’s minimalistapproach to syntactic structure. He recently edited (with Louise Anthony)the volume Chomsky and His Critics and is the author of Move! A MinimalistTheory of Construal.

howard lasnik is Board of Trustees Distinguished Professor Emeritus,University of Connecticut, and Distinguished University Professor, Univer-sity of Maryland. He has co-authored several works in linguistics with NoamChomsky and is the sole author of several books, including the recent Mini-malist Investigations in Linguistic Theory.

david lightfoot is Professor of Linguistics and Dean of the GraduateSchool of Arts and Sciences at Georgetown University. His primary researchinterests are in language acquisition, language change, and syntactic theory;he is the author of eight books, including Syntactic Effects of MorphologicalChange and, with S. Anderson, The Language Organ.

james mcgilvray is Associate Professor of Philosophy at McGill Univer-sity. He focuses on the natures of language and mind, with particular atten-tion to Chomsky’s contribution to these areas; he is the author of Chomsky:Language, Mind, and Politics.

laura-ann petitto is Professor of Psychology and Chairman of theDepartment of Education at Dartmouth College. She is the author of manyarticles on language acquisition and development in children, especially bilin-guals and users of sign.

paul pietroski is Professor of Philosophy and Professor of Linguistics atthe University of Maryland College Park. He is currently interested in theways in which linguistic structure contributes to linguistic meanings. He isthe author of Causing Actions and the forthcoming Events and SemanticArchitecture.

milan rai is a journalist and political activist who lives in East Sussex, UK.He is the author of Chomsky’s Politics and, with Chomsky, of War Plan Iraq:Ten Reasons Against War on Iraq. He recently published Regime Unchanged:Why the War in Iraq Changed Nothing.

carol rovane is Director of Graduate Studies and Professor of Philosophyat Columbia University. She has a special interest in the history of philoso-phies of mind, focusing on Descartes, Kant, and the pragmatists. She is theauthor of The Bounds of Agency: An Essay in Revisionary Metaphysics.


The contributors ix

neil smith is Professor and Head of Linguistics at University CollegeLondon. He focuses on language acquisition and the general linguistic theory,but especially on the work of Noam Chomsky; he is the author of Chomsky:Ideas and Ideals, and, recently, Language, Bananas, and Bonobos: LinguisticProblems, Puzzles, and Polemics.

james wilson is James A. Thomas Distinguished Professor of Law,Cleveland-Marshall College, Cleveland State University. He does researchon imperialism and the use of power, and on the political thought of NoamChomsky. He is the author of The Imperial Republic: A Structural History ofAmerican Constitutionalism From the Colonial Era to the Beginning of theTwentieth Century.



Introduction

James McGilvray

At the time of writing, Noam Chomsky has produced over eighty books,hundreds of articles, and thousands of speeches. He has given thousands ofinterviews, written countless letters, and supervised scores of theses. He hasmade important, sometimes groundbreaking, contributions to three areas –linguistics, philosophy of mind and human nature, and politics. He set lin-guistics on a successful naturalistic, biologically oriented scientific course; histheoretical contributions continue to lead the field. Like Descartes, Galileo,and Hume, and unlike the eighteenth-century philosopher Kant and the greatmajority of philosophers thereafter, Chomsky is both scientist and philosopher,and his philosophical work is continuous with his scientific. His science oflanguage and incipient science of mind offer a genuine prospect of comingto a biologically based grasp of human nature and of the way it allows forhuman understanding and action. His political work, like both Hobbes’s andRousseau’s, seeks a foundation in a science of human nature, although with bet-ter prospects for developing such a theory – and for exploring its implicationsfor political ideals and goals – than Hobbes’s misguided attempt to constructa causal theory of human action or Rousseau’s fanciful assays into a “state ofnature.” And unlike both of them – and far too many contemporary political“theorists” – there is no sign in Chomsky’s political work that his views andcritical analyses are driven by a wish for power.

One purpose of this volume is to offer to a general audience several people’sperspectives on Chomsky’s contributions in linguistics, philosophy of mindand human nature, and politics. The first chapter in each section provides anoverview of Chomsky’s views in these areas. Succeeding chapters developmajor themes. I sketch some of those themes and how contributors developthem near the end of this introduction. A sketch suffices: the chapters andorganization are self-explanatory.

Chomsky the scientist of language

Linguists in the Chomskyan tradition think of themselves as natural scientists –not social scientists, and not engineers. It is important to see what this implies.

1


2 James McGilvray

Ordinary usage is little help. The term “science,” like “language,” has no uniqueuse in everyday speech: people apply it to everything from physics to astrology.And – in part because being a scientist is associated with expertise, specializedknowledge, intelligence, etc. – the desire for social status and political authorityleads to applying the label “scientist” to some questionable candidates. Giventhis, we cannot expect more than a few hints about what science for the Chom-skyan is by looking to the practices that have been called scientific, or to therange of people who have been called (or call themselves) scientists.

A more reliable source is the history of science and the shapes of the sciencesthat are universally agreed to be successful – physics, chemistry, biology . . .Their subject matters and degrees of progress differ, as do their principles, exper-imental techniques, and outstanding problems. There are, however, enough sim-ilarities to draw a composite sketch, especially where the characteristics chosenagree with those from other reliable sources.

Another such source is what those who began the development of successfulsciences said they were doing. Chomsky often mentions Galileo and Descartesin this regard (e.g. 2002b); he considers himself to be working in a tradition ofphilosopher-scientists that they began. These pioneers developed and appliedrecommendations for how to proceed in carrying out investigations of naturalphenomena that led to what were for their times remarkable successes. Focusingon Descartes in his Discourse – a work that explains how Descartes came tohis scientific principles, says what they are, and outlines what he accomplishedby using them – it is striking that he divorces science from another kind ofunderstanding of the world. No one uses scientific concepts in solving themyriad problems encountered in everyday life. Everyone, including the youngchild and the scientist, has and uses what Descartes called “bon sens” (“goodsense”), a practical form of problem-solving capacity that Descartes considersinnate – a gift from God. “Bon sens” is sometimes translated as “commonsense,” and I will adopt that term. It is a capacity to deal with the problemsof politics and commerce, doing the laundry, consoling a grieving friend, andputting out the dog.

While everyone relies on common sense, it does not assume a single formfor all times and all circumstances. This is a benefit. Practical problem-solvingmust accommodate differences in method and individual style, different envi-ronments, cultures, and social organizations, and so on. To do so, common sensemust rely on rich and productive native (innate) resources and a flexible formof mental organization. That is how it can arise so early in children and be soremarkably adaptable. Where Descartes said that bon sens is a gift from God,we are likely to say that its rich resources are biologically based. It is only thusthat we – even the very young – can so quickly conceive, anticipate, and adaptto different environments, adopt (even if only in play) different social roles, andquickly change to meet unanticipated contingencies.


Introduction 3

The contrast to science is instructive. Science is an intellectual projectthat – where successful at all – uses formal (mathematical) theory-constructiontechniques to focus on specific domains; it is guided by a desire for simplic-ity and – as Galileo and Descartes’s work shows – it places simplicity before“data”; it makes progress (often in jumps) over centuries through contributionsfrom many people towards solutions to the theoretical problems it constantlyconfronts, revises, and refines; and it creates its own standards of intelligibil-ity (Chomsky 2002b: 68) that are far from the practical concerns of commonsense. Physics, for example, has taken centuries to develop, has advanced inspurts, uses mathematical techniques to describe a world populated by entitiesand processes beyond the ken of commonsense understanding; and while nodoubt far from complete, it has obviously progressed well beyond Galileo’sand Descartes’s “mechanical philosophy.” This punctuated but deliberate paceis probably necessary because science does not rely on the rich and productivenative systems, flexibly organized, that common sense utilizes. While it can andobviously does rely on apparently innate senses of simplicity and what countsas a good explanation and description (Chomsky 1980), construction of thetheories that solve the problems science confronts requires invention, favorableconditions, and cooperative activity. That is why – with the exception of parts ofmathematics – only rudimentary forms of natural science developed before theend of the sixteenth century. Unsurprisingly, it also takes a considerable amountof time and training for individuals to acquire sophistication even in a specificscience; the full range of the developed sciences is out of the reach of everyone.Fortunately, science’s findings are not needed for survival, or even to thrive.No doubt doing laundry benefits from engineering applications of fundamentalscientific principles – those that lead to variable-speed electric motors and front-loading washing machines. But for millennia people managed with technologythat required only the engineering solutions offered by unaided common sense.They built bridges of various materials, annealed metals into Samurai swords,and constructed cathedrals. In sum, science brings the developed formal toolsof highly focused inquiry to bear on theoretical problems; progress – relyingas it does on invention – is usually slow. Using their common sense, peopleutilize native resources, perhaps in forms of practice that have led to practicalsuccess before, to deal with the immediate demands of everyday problems. Weinvent scientific tools to deal with bosons and genomes; we depend on nativeresources to critically assess the performance of an elected representative – orthe intentions of an artist.

One of the characteristics of scientific practice that science’s history revealsis seeking a particular kind of objectivity – one that is universalized, so that it isnot tied to person, circumstance, culture, or history. That notion of objectivitycannot serve the tasks that common sense deals with; commonsense understand-ing’s concepts are “designed” to serve matters of human interest – including


4 James McGilvray

those of perception and action. It should be no surprise, then, that sciencecan lead to denials of the “obvious” claims of commonsense understanding.Sciences of the mind tell us that the colors we experience are products of ourvisual systems, not “on” things outside, and that languages – including theirsounds and meanings – are native and in the head, not somehow outside thehead, perhaps products and properties of communities and polities. Scientistsoften must ignore appearances (as in colors and words outside the head) andinvent, using the tools that mathematics – much of which is invented too –provides. And they must measure progress not by how well a proposed changein a theory satisfies untutored opinion or “raw experience” but by improvementsin description and explanation of the relevant phenomena and greater formalsimplicity. Descartes – had he lived long enough – would have seen his contactmechanics refuted by Newton’s gravitational principle. The “obvious” idea thataction and effect require contact seems to have its origin in common sense; itfails in science. Physicists learned long ago that the apparently obvious is atbest a starting point. That lesson has been hard to learn with language, as wewill see.

Chomsky’s science of language is a science in the Cartesian–Galilean tradi-tion. It is a branch of the study of biology. It is a naturalistic science that providesan “abstract” description and explanation of a biological system found only inhumans, the system that Chomsky calls “the language organ.” The languageorgan revealed by Chomskyan science of linguistics is far from the common-sense idea of a language as a social phenomenon. To reveal this organ, thescience of linguistics had to develop standards of intelligibility that were con-sonant with those of the natural sciences, not with what some philosophers call“granny’s view” of language. The result, after several decades of work, is thatthe language organ appears to be remarkably simple in its “design.” This isunusual in biology, a domain that usually reveals what Jacques Monod callsthe “tinkering” of evolution. Apparently, extending naturalistic science to thestudy of a biological system of the mind yields a fascinating result: languageconfirms Galileo’s and Descartes’s vision of a well-designed, elegant nature.

Descartes’s mechanics – even in its rather primitive form – conflicted withthe “obvious” principles taught by church and universities. It conflicted withthe teleological world of Aristotle, modified by the seventeenth century tosuit Christian doctrine. That is why his and Galileo’s novel mathematical–mechanical theories of natural phenomena faced opposition from philosophicaland religious systems that, like many today, take their task to be that of defendingideas that have their origin in common sense with its practical, not theoreticalorientation. Successful sciences since Galileo and Descartes have continuedto use simple, elegant, formal mathematical tools and invented theories andconcepts to provide descriptively and explanatorily adequate theories of theirdomains. Like Galileo’s and Descartes’s sciences, they continue to be opposed,although some of the opposition is muted by the obvious success of the theories.


Introduction 5

Chomsky’s naturalistic inquiry into language, like Descartes’s into cosmol-ogy, physics, optics, and neurophysiology, also gets opposition from the experts.Opposition comes from several fronts, but most seem to proceed on the assump-tion that language is not a natural phenomenon. They see language in terms ofits use – perhaps as a set of social practices, a bunch of “tools” we have made tocommunicate, etc. In each case, one finds a version of what Chomsky calls anE-language approach to theorizing about language (external approach). Amongthe majority of philosophers, it appears as insistence that one or another formof the “obvious” idea that language is an institution created by humans to com-municate – a “practice,” a product of history, a set of habits, an “interpretivemedium,” a mode of communicating a speaker’s intentions. From psychol-ogists, philosophers, and other cognitive scientists wedded to one version oranother of what Steven Pinker (2002) calls the “blank slate” picture of the mind,it comes as the idea that it is a form of behavior that solves cognitive problemssuch as classifying, describing, etc. Any organism or device that displays the“same” behavior has a language, they believe, and they “get” it by whatevermeans (training, programming) the blank slate advocate employs. So getting anape (on which see Petitto, this volume) or a machine to simulate the behavior“proves” that language is not a biological organ with which only humans areborn.

Perhaps these experts succeed sometimes by some standards – it is not clearwhich – but not by those of naturalistic inquiry. Philosophers who think languageis a social phenomenon that children learn from their community ignore the factthat languages are quickly acquired by very young children without training.The same is true of concepts of social role. If neither language nor conceptsof social role (and much else) is taught, they must somehow be built into thechild’s mind at birth; that is where to focus naturalistic inquiry. And blankslate advocates need to learn the elemental lesson that it is unwise to focusattention on sameness of behavior or “output” and the means by which theseare induced. Even if one succeeds at getting a machine or ape to “speak” –it has not happened, probably for reasons that Descartes (and Alan Turing)pointed to – that is no proof that the systems that make this behavior possible(which is where Chomsky focuses his work) are the same as the ape’s or themachine’s. Manufacturing an excavator to dig ditches hardly proves that humangravediggers moving shovels with their articulated arms have hydraulic systemsin their arms activated by diesel-powered compressors.

Chomsky on biology and evolution

To avoid confusion that might arise from speaking of Chomsky’s view thatlanguage is a biological organ, I need to mention the matter of evolution (fordetail, see Jenkins 2000). Chomsky, like Richard Lewontin (1990), has littlesympathy for current efforts (e.g. Pinker & Bloom 1990) to try to show that


6 James McGilvray

language – especially in the form of the basic computational system that linkssounds and meaning to produce a discrete infinity of sentential expressions –is the product of some sort of natural selection that tracks increased repro-ductive advantages afforded to those who are (on the Pinker–Bloom story)better communicators. Chomsky does not doubt that language evolved, in somesense: it is biologically based and appeared in the human species. And he hasno doubt that it has proven to be extremely useful to humans. But selection-for-communication, selection-for-some-function-or-another, and even selec-tion, period, do not exhaust the field. Pace Pinker and Bloom, there arealternatives.

One problem with the attempt to show that language was selected by repro-ductive advantage is that humans as a species have been relatively stable fora long time – probably 100,000 to 200,000 years. So language in the formof the basic computational system that seems to be unique to us must haveemerged somewhere between now and 100,000 to 200,000 years ago. (Thebest current guess is approximately 60,000 years when the migration fromAfrica began.) It is all but impossible to find evidence in observable phenom-ena for the selectional emergence of such a system. Perhaps we will somedayidentify the computational-system specific gene(s) that provides us our lan-guages’ syntax; perhaps too by investigating remains we could say when thisgene was introduced. But nothing would tell us why and how it developed.Speculation about selection-for-any-function of language’s computationalsystem (its “syntax”) seems to be empty.

But that is not all that needs to be said. For one thing, while looking for a his-torical record seems hopeless, we can compare. Hauser, Chomsky, and Fitch dojust that in their article in Science (2002). Comparison offers no immediate helpto the Pinker–Bloom selection-for-communication cause, however: language’sbasic computational system can produce a discrete infinity of sentences, andthere is no current evidence that other species can “express” discrete infinitiesof elements of any sort. For example, no other species enumerates arbitrarilylarge numbers of elements in a set by counting them out. But perhaps we havenot looked far enough. Most comparative studies focus on human and animalcommunication systems; perhaps the computational system is found elsewhere.In concluding their discussion, Chomsky and his co-authors suggest looking atother kinds of system: look, they suggest, for non-communicative systems thatrely on a recursive computational procedure that provides for a discrete infinity.Perhaps they will be found in navigational systems, or those that “parse” socialrelationships. If there were such a system, it – or its homologue or analogue inthe developing human species – might have been exapted (coopting a systemadapted to serve another function) for language. Selection-for-communicationwould fail, but perhaps aspects of the selectional cause would be salvaged. It isa project worth trying.


Introduction 7

In his own work, Chomsky suggests we look wider still. Darwin pointed outthat selection is only one part of evolution. And there are other, non-selectionaltraditions of biological development and speciation. One such tradition is foundin Stephen Jay Gould’s and Richard Lewontin’s suggestion (1979) concern-ing “spandrels” – structural consequences of other, perhaps selected, systems.Another tradition, perhaps related, goes back to Goethe and his discovery ofan Urform for plant morphology (cp. Chomsky 2002a: 66). Goethe thought hehad discovered a formula that predicted all (biologically) possible forms plantscould take. If there is such a formula, it indicates that there lies in plant morpho-genesis a physical factor that yields very different-appearing plant forms (and“species”), given slightly different “input” conditions. The formula and dif-ferent physical conditions, not selection, would account for differences. Thattradition was represented in the nineteenth century by several individuals inEurope and, in the twentieth century, versions of it appeared in mathematicalform in the work of D’Arcy Thompson (1917) and Alan Turing (1952). Manyhave pursued their suggestions; there is a growing mathematical science ofmorphogenesis.

Chomsky sometimes suggests (2002b: 57) that the development of the lan-guage organ might be explained as a mathematical consequence of the kind ofcomplex form of mental biology humans have. Even selection has to operatewithin the “channels” provided by basic physical processes, after all; and ourlanguage faculty appears to be too “perfect” a solution to linking sounds andmeanings to be the result of selectional tinkering. Perhaps the computationalsystem built into our languages is “anticipated” in those physical processes andstructures and arises “by itself” when other systems are in place. This wouldallow that the computational system of language came about as a completepackage, perhaps 60,000 years ago. Or perhaps language is a spandrel. Eitherway, we abandon the gradualist, “historical” form of development and biologi-cal differentiation (and the “tinkering”) that the selectional picture relies upon.We might even be able to find evidence.

In sum, while there is no doubt that language has proven to be an extremelyuseful biological faculty that has given the human species extraordinary cog-nitive advantages, there is little reason now – and there may never be – to holdthat the computational core of the language organ developed slowly over a longhistorical period by virtue of affording reproductive advantages to successivegenerations of communicators.

On the unity of Chomsky’s thought

A person’s intellectual work as a scientist need not be connected to his or herpolitical views – there is no reason, for instance, that a biochemist’s scientificwork should have anything to do with her neoliberal views. But Chomsky’s


8 James McGilvray

linguistics and his political views seem to be special cases, particularly whenone takes into account his philosophical/scientific work on the human mind andhuman nature.

One reason to look for connections and perhaps even a degree of convergencein all three areas of Chomsky’s work is that each has, in its own way, somethingto say about human beings. More narrowly, each focuses on distinctive featuresof human beings – on language, a biologically unique mental faculty; on ourdistinctive natures and minds with their limited but biologically unparalleledintellectual capacities for dealing with both practical and scientific problems;and on those apparently unique forms of social organization that we think ofvariously as polities, communities, societies, and/or cultures. No other organismcreates for itself organized groups of non-kin individuals in ways that allow forcooperative, non-contact, coordinated ways to meet needs and solve problems.

Some non-Chomskyans think that there is a connection between languageand culture or community. Philosophers as varied as Foucault and Putnam andpsychologists as different as Piaget and connectionists share the assumptionthat language depends on the society, culture, community (etc.) in which oneis raised. By “depends” I mean not just that children born into a group ofJapanese speakers come to speak Japanese; everyone acknowledges that andtries to explain why it happens. Rather, they take language to be constitutedby the society or community in which one is born. This idea often appearsas the view that children are taught the language of their community by theirelders: the elders know the “rules of correct usage/correct practice” (givenrelevant circumstances) and instruct by encouraging correct verbal responsesand discouraging incorrect. People (as a group), over a long historical period, arebelieved to have invented the practices that define a specific community, culture,etc. and, while doing so (or perhaps in doing so), to have also invented language –another form of practice that happens to allow individuals to communicate,coordinate, etc. Individual creativity in the exercise of one’s intellectual powersdoes not figure in this story; the focus is on community practices/habits/rules forapplying words correctly, etc. If defenders of this idea speak of human natureat all, they make it a historically conditioned notion: as people’s fundamentalpractices change, people’s social/cultural “natures” change. Alternatively, theymight say that human beings are plastic (people are intellectual/cultural blankslates), so that human social/cultural natures are – unlike those of any otherbiological species – molded by the societies, cultures, etc. in which they areborn.

Chomsky’s view of language and the mind reverses priorities. For him, humanlanguages are not expressions of culture and society – in effect, human arti-facts. They are, in a sense, expressions of our genes: all the existing and possiblenatural languages (not technical symbol systems, such as those found in the sci-ences) are biologically encompassed within what he calls “Universal Grammar.”If there is any dependency between language so conceived and society, culture,


Introduction 9

etc., it cannot make culture the condition of language. If anything, culture (etc.)depends on language. Suggesting that culture depends on language in this senseis not making a causal or deductive claim. The language organ does not secretecultures or social arrangements. Rather, language provides the rich, unlimitedset of conceptual structures (Chomsky informally calls them “perspectives”)and the opportunity to communicate them that humans need to conceive ofalternative ways to solve the problem of how to live together to the benefit ofall, to discuss and come to agreement on the options, and the like. In effect,language and our other cognitive resources, but especially language, make itpossible to create cultures and much else. Adopting this point of view – thatnative conceptual tools, and especially language, must be in place before artic-ulated conceptualization and understanding, much less discussion, can occur –another matter falls into place too. Individual creativity – a curiosity on theculture-first approach – can now be seen as benefiting from the infinite scopeof linguistic output of which our systems are, in principle, capable.

To see why language should have a central role in making sense of howwe come to create our diverse communities and cultures – and individual cog-nitive and expressive styles – it is important to keep in mind that humansare the sole species to have language. Many other species have communica-tion systems. And some others also have the “performance” systems that areinvolved in human language: auditory perception and production (for speech),visual perception and aspects of articulatory shaping (for sign), plus aspectsof those resources that Chomsky calls “conceptual and intentional” – thosenon-linguistic resources that can be brought to bear on circumstances to yieldvarious forms of intelligent behavior. But no other species has the capacity todevelop a potentially infinite, discrete set of mental “outputs” in the form ofexpressions or sentences that link perception-related configurations, whethersound or sign, with conceptual materials (Hauser, Chomsky & Fitch 2002).That is, no other species can produce – apparently at will – innumerable sets ofsentences or expressions. Given the obviously central role of language in humanthought and action, our distinctive mental capacities – found in both practicaland theoretical problem-solving – may be due, in large measure, to language.And, with these capacities, we also can develop social organizations: we canplan, organize, decide to cooperate, and create institutions. It becomes quiteplausible that culture and our various forms of social organization depend onlanguage rather than the other way around. So we have one connection betweenthe areas Chomsky works on: the science of language might well provide thekey to what is distinctive to our minds and natures, to making sense of why wehave the distinctive mental capacities we do and, in turn, making sense of howwe can create our various forms of social organization.

Another kind of connection depends on the fact that views of human nature arealways behind people’s attempts to justify their moral and political principles.In the background of every political and moral “ism” (including those largely


10 James McGilvray

indistinguishable forms of corporation-dominated plutocracy–oligarchy called“neoliberalism” and “neoconservatism”) one finds assumptions about humannature – about what human beings “are” and what they are or are not capable of.These views of human nature typically play a justificatory role. “That’s a sillyview of democracy,” someone might say of the fully participatory form Chom-sky favors, “people (of their natures) don’t have enough interest, intelligence,knowledge, or time to participate fully. We must give an elite managerial classthe power to make decisions and run the economy, government, courts . . .”Someone else might say: “People are naturally aggressive and acquisitive. Agood form of government must have full authority to restrict their uncheckedexercise (a Hobbesian state of nature); we need authoritarian government toprovide a form of rescue.”

While justification of this sort is common, few of those employing it botherto elaborate their view of human nature. And the connections between whateverdegree of articulation one finds and the moral/political/religious . . . claims theyare supposed to justify can be quite hazy. Moreover, there is little if any effortto show that one’s view of human nature is itself justified by the standards ofempirical inquiry. A biologically based science of human nature would avoidthese problems. Appeals to gods and revelation, or to what seems to be obviousto some group or another, are almost always self-serving efforts which reveala desire to place or maintain oneself or one’s group in a position of power orauthority. We need a detailed, objective view of human nature, and scientificinquiry can provide that. It alone can say what is distinctive about our natures –as opposed, say, to those of various other primate species.

A plausible way to focus such an inquiry would be to look for aspects of thehuman mind that are distinctive – for faculties or forms of mental organizationthat humans have that other primates lack. The faculty of language, clearly,is such. A science of language and of what language provides humans shouldthus have an important role in such a science. Not only does language seemto be unique to humans, but it also seems to contribute to a unique form ofmental organization. The biological faculty/organ of language acts rather like acentral cognitive system, allowing us to coordinate materials provided by othercognitive systems in ways that other creatures seem to be unable to manage.And, of course, it provides the conceptual tools to allow us to speak of anything,anytime. In these and other ways, it enables us to “solve problems” in a widevariety of manners. It is almost as if language allowed our minds to be “universalinstruments,” to use Descartes’s Discourse terminology. So Chomsky’s scienceof language, even in its incomplete state, represents a good beginning to ascience of the human mind and, thus, of human nature.

While the science of human nature is in its earliest stages, we can use whatwe have now to begin to think about how to craft a good society. We cannotmove directly from biology – specifically, the biology of the human mind and


Introduction 11

what it provides us (for these are what make us distinctive in ways that weso obviously care about) – to a picture of an ideal. We must start by decidingwhat an ideal society should accomplish – what its function(s) should be. Forsocial organizations are institutions, made by human beings, and a good oneshould fulfill its function well. A plausible suggestion is that the function ofa human society is to meet not just the needs of survival, but those that arecharacteristic of the kinds of creatures we are. Call these characteristic needsof humans “fundamental distinctively human needs.” Now what the science ofmind tells us about language and the rest of the mind, and about how people usethese cognitive tools, can come into play. To be brief, because we have langu-age and language seems to be the key to our extraordinary mental capacities, wealone seem virtually designed to be creative creatures. Our languages providean unlimited range of “perspectives” (Chomsky 2000a: 150, 180), and these canbe – and are – used to serve all sorts of purposes, including those of art and labor.Language’s unlimited range comes to play a role in virtually all our affairs – notjust our thoughts and efforts to understand others, but in our jobs and everydaytasks, even putting out the dog. An ideal form of social organization must, then,give individuals ample opportunity to exercise their creativity. This need notmean that we must all become craftsmen and painters or composers. It mightmean that if we labor with others in a factory, we have sufficient freedom andopportunity to fully contribute to all decisions that concern us, to bring aboutchange, and to otherwise control the conditions under which we work. Or itmight mean that operating an excavator, we not only do the job well and witha concern for those who will use what we do, but with a form of artistry.

Another candidate for a fundamental need is that for community, friendship,love, and nurture. It is only if one thinks of this need as that of an animal thatis also “bred” for freedom and creativity, though, that the need for communitybecomes distinctively human. Many other primates display in their behaviorsa need for association for mutual benefit. But the range of options available tothem and the forms of organization that they can conceive are very much morelimited than ours. They do not seem to be able to conceive alternatives, choose,and plan. Their “institutions” are suited to specific environments, with specificforms of threat and opportunity. They do not seem to be “made.” And there islittle change in them.

The experiments of history lead to a similar conclusion about fundamental,distinctively human needs: people need to be free and create while integratingthis with ways of associating with others. Where we find people willing tomake considerable sacrifices for goals other than mere survival (sacrificing forthe young and future generations, revolting against oppressive authority, reject-ing slavery – Chomsky 1988b), we can plausibly assume that the goals of thesacrifice represent fundamental needs – trumping even those of survival. Historyalso suggests which forms of organization best meet needs. It reveals people in


12 James McGilvray

various forms of social organization in various environments. Investigating, wefind cases where people resist the forms of organization they find themselvesin and aim to improve them. Equally important, we look in the directions thatthey seek improvement, and we note the success of the solutions they work out.And where there is progress in meeting these goals in the social organizationsthat develop, we note that. There are complicating factors – as always, wherethere is entrenched power. And the tools of power have become increasinglysophisticated, especially in the twentieth century: media control, the tools ofadvertising, and similar forms of information control have proven to be pow-erful forms of mind control. But history suggests that people need choice andautonomy, and to place their stamp on the work they do.

Chomsky suggests that an ideal form of social organization would be oneor another form of what he calls “anarchosyndicalism” or “libertarian social-ism.” These are “isms” that one does not usually encounter, although theyare suited to the idea that freedom and community can integrate. The anar-chist/libertarian aspect would satisfy the need for freedom and creativity, andthe syndicalist/socialist that for community (often found not just with family,but with those at work, in community projects, at play, and so on). Explanationsof why Chomsky chooses this as an ideal form of organization (and insight intohow this political ideal guides his criticism of current political “management”and suggestions concerning policy) are found in his political writings. I do notpursue the matter further here. My aim is to indicate that seeking justificationfor a moral/political ideal represents another kind of connection between a sci-ence of language and politics. Again, the empirical scientific study of mind(prominently, language) and human nature plays a central role.

Whether readers pursue the question of integration or not, I hope this volumewill encourage all to look further into Chomsky’s work and the work of thosewho have extended it – including those in this volume who, in discussingChomsky, speak not just of his work, but of what they and others have beenable to contribute to “Chomskyan thought.” Their efforts illustrate how fruitfulChomsky’s contributions have been to the intellectual study of language, ofhuman mind and human nature, and of our conduct and goals as members ofpolitical communities.

The chapters

Linguistics

Chomsky’s work in the science of language began in the late 1940s with anundergraduate thesis at the University of Pennsylvania, the basis of his MAthesis, The Morphophonemics of Modern Hebrew. After appointment as aHarvard Junior Fellow in the early 1950s, he began the monumental The Log-ical Structure of Linguistic Theory, a chapter of which was submitted as his


Introduction 13

Ph.D. thesis at Pennsylvania. Completed in 1955 and revised for publication in1956, it was not actually published until 1975, and then only in part. But, asHoward Lasnik mentions in his discussion of the computational “levels” builtinto Chomsky’s various theories of the language organ, it set the stage for, andanticipated, much of Chomsky’s work in the science of language, includingaspects of his recent “minimalist” program.

Neil Smith nicely outlines the nature of Chomsky’s project for a science oflanguage. He also points to the connection between the science of language andbiology. That connection becomes particularly evident in Chomsky’s solutionto what he calls “Plato’s Problem” – the task of explaining how we can acquireso much knowledge of language (its structure, sounds, and meanings) in sucha short time. David Lightfoot focuses on this. The solution Chomsky offers,Universal Grammar (UG), is a hypothesis about what children start with, the“initial position.” Lightfoot looks primarily at what must be a biologicallyinbuilt, structural schema for language. Succeeding chapters discuss otheraspects of UG. Elan Dresher focuses on Chomsky’s work on linguistic soundsduring the 1950s and 1960s, culminating in Chomsky’s and Morris Halle’sSound Pattern of English. This work indicates that human linguistic soundsare systematic, “abstract,” and, apparently, unique to language alone. Furtherdeveloping a small but revealing segment of this theme, Laura Petitto discussesresearch that localizes tissue in a part of the brain homologous to that found inseveral primates, tissue that used to be thought of as devoted to sound recog-nition but that in the case of humans seems to be language-specific, innately“programmed” to recognize, respond to, and lead to production of linguisticsyllabic structure – syllabic structure, remarkably, in both speech and sign. Thelast chapter of the linguistics section presents some of Lila Gleitman’s andCynthia Fisher’s work on “word” (lexical) learning. They do not attempt to saywhere sounds and meanings (concepts) come from. Presumably, a full theoryof UG speaks to that. Instead, they focus on the kinds of information childrenrely on in order to associate or map sounds and concepts in their mental diction-aries. That information is syntactic and language-specific, which presupposesthat the child, who so obviously recognizes what is relevant (and when), hasthe conceptual tools and a schedule for their application built into the mindat birth. A theory of UG describes those tools and points in the direction ofmaking sense of how they come to be applied.

Philosophy of mind

Like Descartes, Chomsky makes his philosophical work continuous with hisscientific. The study of mind and language is an attempt to characterize andexplore the consequences of a developing science of mind – one in which ascience of language plays a prominent role. The consequences that can andshould be explored include those that involve the realization that we – and our


14 James McGilvray

language – are biological structures. This suggests a different way to under-stand ourselves and human communities and, thus, consequences for action –including political action. Seen in this way, philosophy well practiced aims tooffer the best – most rationally defensible, all things considered – picture of ourbiological minds and natures while exploring the consequences if the picturethat is drawn is correct.

Chomsky’s work in this area began in the 1950s when he read historical worksin linguistics and philosophy. His reading led him to develop and elaborate aframework for understanding the human mind and language that he called a“rationalist” approach, which he contrasted with a rival “empiricist” approach(Hornstein, this volume). The labels and the ways Chomsky characterized therationalist and empiricist pictures of the mind are apt: they suit the philosophicalviews of the mind that traditional rationalists (Descartes, Leibniz, Cudworth,etc.)1 and empiricists (Locke, Hume, etc.) held, and they characterize importantdifferences between views of the mind and the science of mind found today.The contrasting approaches are explored in detail, although not under theselabels, in Cartesian Linguistics in 1966 and they have been elaborated since.Rationalists hold that the mind is both structured and provided with rich andextremely useful “content” at birth. Rationalists are “nativists.” The rationalistrecognizes, of course, that experience and “external” factors play a role in themind’s “choosing” which concepts to activate or develop. But the rationalistdenies that external elements shape and constitute concepts via the operations ofsome sort of domain-general learning procedure such as hypothesis formationand testing. Circumstances serve to “occasion” or “trigger” the introduction ofa concept; crucially, the mind’s own machinery dictates what “patterns” in thedata count as appropriate “occasions.” The patterns are, in a sense, built intothe mind all along. Empiricists, in contrast, think of the human mind as gettingits language-specific (and much else) structure and virtually all of its “con-tent” by “learning” it from environmental conditions – interaction with things(world) and others (“speakers” who “train” their young, according to some).Chomsky’s empiricists are anti-nativist. Further, his rationalists think that themost fruitful way to study the mind and its elements and “contents” is to focuson its internal structure and operations (“internalism”). For them, a study oflanguage’s various contributions to cognition, including the concepts that lan-guage expresses and the terms it puts in “referring positions” in sentences, is astudy of internally constituted “tools” that people can use for various purposes.The ways people use these word-tools to say what they intend, or to refer tovarious things, are matters of free human action. Trying to deal with free actionin a naturalistic science is for the rationalist hopeless. In contrast, the empiricistwhen dealing with concepts or reference looks outside to the mind/person inits/his/her interaction with the environment (“externalism”). They might con-strue a concept as a functional role in some overall account of humans using


Introduction 15

language (Wittgenstein 1963; Sellars 1974). And they might think of referenceas a conventional relationship between word and world, or perhaps even lookfor a “natural” relationship in a realist construal of information theory (Dretske1981); in either case, they apparently ignore the fact that people use words (andthey use them in many ways) to refer to many things.2 In some much-admiredwork (Kripke 1972; Putnam 1975; Burge 1979), the idea that environment con-stitutes linguistically expressed concepts/meanings has taken the extraordinaryform of holding that meanings are individuated by properties and things out-side the head, perhaps completely unknown to speakers. I might be told thatthe word platypus means – at least in part – some specific set of features of theplatypus genome; these are unknown to me and – I venture – to anyone else.Yet that genetic structure is part of what I express when I say “Doesn’t Harriethave a pet platypus?”3 Finally, Chomsky’s rationalists attribute the remarkableflexibility and adaptability of humans and the creativity of the human mind tothose internal, largely innate structures and contents of the mind that enableflexibility, adaptability, and creativity. His empiricists, committed by the natureof the empiricist program to trying to find system in the multiple ways in whichlanguage and other cognitive systems are used, have little to offer in explaininghuman creativity. They tend to gesture in the direction of similarity and anal-ogy, and say that these extend already-learned structures and contents. Chomskyindicates the failures of this approach in many places (1966, 1975, 1986, 1988b,2000a, inter alia).

Norbert Hornstein outlines Chomsky’s rationalist/empiricist distinction inboth its historical and contemporary forms. Paul Pietroski and Stephen Crainpresent – focusing on recent evidence – a discussion of the nativist/anti-nativist issue. They explain why Chomsky thinks there are “innate ideas.” AkeelBilgrami and Carol Rovane show why Chomsky thinks that the current philo-sophical – and dominantly empiricist – preoccupation with language–worldrelationships under the topic “reference” is misguided. They also outline someaspects of Chomsky’s view that human understanding and knowledge – basedon biologically native systems as he thinks it must be – is of its nature limitedby the cognitive “equipment” with which biology has provided us. In the lastchapter of this section, I outline some of the considerations that lead Chomskyand several other rationalists to the conclusion that much of our conceptualrange is built into us at birth. Following Chomsky (1966, 2002a), this idea isthen linked to human creativity. That in turn introduces an important theme ofChomsky’s political views.

Politics

Where Chomsky’s elementary school classmates might have turned their com-monsense form of understanding to analyzing sport teams or to detailed analyses


16 James McGilvray

of who was friends with whom, he focused on politics. His first politicalpublication, a February 1939 reflection on the fall of Barcelona, Czechoslo-vakia, and Austria and the ominous rise of fascism, appeared soon after histenth birthday in the school paper he edited. It helped spark an interest in theSpanish Revolution, a topic he could pursue in his trips to New York, wherehe frequented anarchist offices and the secondhand bookstores on 4th Avenuerun by refugees from fascism (see Barsky 1997 for an account of Chomsky’searly life). The research he did in his teens – plus, no doubt, a continued intenseinterest in the creative and developmental opportunities afforded individualsby a rich native endowment, including the possibility of anarchist forms ofsocial organization – allowed him, many years later, to write a sophisticatedreview of a scholar’s book on the topic. Chomsky has a prodigious memoryand intense powers of concentration; few others could recall or consolidateenough of what they had read to use it many years later. What needs emphasis,though, is that his classmates had the same tools – the concepts of commonsense – needed to understand people as individuals and in groups that Chomskyhad.

The obvious lesson is that political systems and political events are withinthe reach of everyone: “experts” and “managers” are not needed for politicalanalysis, criticism, or decisions. Political study and criticism is not science; thelabel “expert,” warranted where specialized concepts are in play, does not apply.Granted, some people are more sophisticated in political analysis and criticismthan others, and anyone can improve in discernment – experience tempered withskepticism does matter. But sophistication and improved discernment requireno specialized knowledge and arcane concepts, just interest and – connectedwith that, surely – some expectation that one’s interest can make a difference.Given this, why is there often more interest in sport and the latest Elvis Presleysighting than in political analysis and criticism, even in contemporary democ-racies, where – presumably – one can make a difference? Part of the answerlies in the fact that contemporary democracies are largely in the control of pri-vate power – in effect, corporations. Contemporary democracies respect JamesMadison’s principle that those who own the country should run it. Individualvoters choose from a list of candidates, often representing a narrow politi-cal range. While representatives nominally have considerable power and areelected to serve the interests of their constituents, their decisions in fact servethe interests of private economic power. The mechanism is straightforward:corporations won for themselves at the end of the nineteenth century manyof the rights of persons (including free speech), and they use these “rights”and their massive economic power to influence elections (by contributions andadvertisement) and to determine legislation (by lobbying and threat). In thisway, the important economic decisions – those that are so crucial to people’s


Introduction 17

lives – are, in effect, made solely by boards of directors of corporations, institu-tions designed to maximize accumulation (profit) and domination (monopoly),not “serve the people.”4 Most of the electorate recognizes that the systemaccords them little control; that is why there is often little interest in the politi-cal process. These considerations are rarely mentioned and never emphasizedin corporate-run mass media, which studiously overlook the obvious ways inwhich policy, both domestic and foreign, accords with the needs of corporatepower. For example, one of the consequences of continued US hegemony in theMiddle East is a considerable degree of control over the oil resources of thatregion. That hegemony has, in part, been maintained since the mid twentiethcentury by massive financial aid to Israel – particularly in the form of militaryequipment. Other “initiatives” included military support for Turkey, “friend-ship” with the Saudis (who invest heavily in US markets), aid to an Egypt will-ing to accommodate Israel and suppress popular local movements, and supportfor Saddam in Iraq during the 1980s – including support for chemical warfareagainst Iran and a blind eye to Saddam’s efforts to slaughter Kurds. When Sad-dam became less useful and made the mistake of challenging US control, it led toBush the first’s invasion, followed a decade later by Bush the second’s. The mostprominent beneficiaries of these efforts have been not the citizens of Iraq or theUS (who inevitably must pick up the tab), but corporations (e.g. Halliburton) andmarkets, which, assuming US hegemony, are assured of continued control andlow energy costs. The pattern is the same in other cases – even Vietnam, whichhas joined the capitalist market fold. No US administration publicly admitsthe imperialist intentions of the project of “bringing freedom” (i.e. freedomfor markets and corporations) (Chomsky 2003), and corporate-owned mediaseldom mention this or other unwelcome facts about the nature of the project(Chomsky 1989). Instead, both craft and foster what Chomsky calls the “neces-sary illusions” that portray the 2003 invasion of Iraq, for example, as motivatedby an effort to bring democracy to the Iraqis, fight terrorism, corral the Butcherof Baghdad, etc. Chomsky deals with this phenomenon – and its motivations –under the topic “the manufacture of consent.” Jean Bricmont explores this topicin the last chapter of the volume; the other authors in the political section alsotouch on it. Chomsky’s work in this area – often in cooperation with EdwardHerman – has to be counted as one of the most important – and increasinglyinfluential – studies of political behavior in recent times.

Milan Rai provides an overview of Chomsky on politics by placing politicalviews in an Enlightenment conception of morality. Chomsky has spoken to somany political topics and issues that it is impossible to offer a complete picture;but one can – and Rai does – bring many issues together by detailing Chomsky’smoral motivations: why in all his political efforts he emphasizes freedom andreason. The three other contributors focus on central themes in Chomsky’s


18 James McGilvray

political works. Jean Bricmont has been mentioned. James Wilson discussesChomsky’s views of the individual, the state, and corporation, and relationsbetween them; this helps make sense of Chomsky’s views of the United States’(and other corporate-run states’) internal affairs. Irene Gendzier focuses onChomsky’s attempts to elucidate the motivations of US foreign policy: sheconcentrates on the North/South divide and US imperialism.


1 Chomsky’s science of language

Neil Smith

Language makes us human.Whatever we do, language is central to our lives, and the use of language

underpins the study of every other discipline. Understanding language gives usinsight into ourselves and a tool for the investigation of the rest of the universe.Martians and dolphins, bonobos and bees, may be just as intelligent, cute,adept at social organization, and morally worthwhile, but they don’t share ourlanguage, they don’t speak “human.” One of Chomsky’s achievements is tohave demonstrated that, despite the easily observable richness of the world’slanguages, there is really only one human language: that the complex andbewildering array of different languages surrounding us are all variations on asingle theme, most of whose properties are innately given.

The scientific study of language

Linguistics is conventionally defined as the “scientific study of language” –“language” in the singular. Although its domain is usually taken to includenot just the wealth of the world’s languages – Amharic, Berber, Chinese,Dutch, English, etc. – but also all possible languages, past, present, andfuture, for Chomsky the focus of linguistics is the study of knowledge of lan-guage, of “human.” This chapter will attempt to justify and explain this empha-sis, spell out its implications, and motivate the description of his linguistics as“scientific.”

There are several strands to the claim that linguistics is a science. The first isthat linguistics provides a general theory explaining why languages are the waythey are: each language is a particular example of a universal faculty of mind,whose basic properties are innate. The second is that the theory should spawntestable hypotheses about those properties: like a physicist or a biologist, thelinguist manipulates the environment experimentally to see what happens and,crucially, he or she may be wrong. The third is that the investigation of languageshould proceed no differently from the investigation of physical and chemicalentities: there is no justification for placing requirements on linguistic theoriesbeyond those placed on physical theories. In both domains, the theories are

21


22 Neil Smith

underdetermined by the data, but their respective hypotheses are comparable inthat they aim to discover the truth about aspects of the natural world.1

It is important to note that characterizing some enterprise as “science” isnot a value judgment. Dostoevsky’s insights into the human condition are asdeep as those of Darwin, but they have a radically different status. As Chomskyputs it: “We will always learn more about human life and human personalityfrom novels than from scientific psychology.”2 Literature and science are com-plementary rather than in competition, and domains which lend themselves toscientific investigation are few. Vision and knowledge of language are suchareas, consciousness and free will are (probably) not. The former constitute“problems” about which we can devise explanatory theories in much the sameway as we devise theories of particles or genes in the natural sciences; the latterare “mysteries,” whose understanding in any depth may well lie beyond ourintellectual powers (1975: ch. 4). It is plausible to assume that spiders are incap-able of understanding the geometry of the elegant webs they construct. It isequally plausible that we have comparable limits to our understanding, and thatunderstanding free will lies beyond those limits. Those domains about whichwe can construct theories of the kind characteristic of the natural sciences aresaid to fall within the scope of “naturalistic inquiry.” Knowledge of languageis such a domain.

Modularity

Language may be what makes us human, but humans are remarkably complex,and the human mind/brain3 is notoriously the most complex entity known.Fortunately this complexity can be broken down into more manageable chunks,where each chunk constitutes an appropriate domain of investigation and the-ory construction. While they can be related, physics and chemistry, botany andzoology are distinct domains; and within any one of them there are finer sub-divisions, so that respiration and reproduction, or the circulatory system andthe olfactory system are treated separately. This attempt to divide and conqueris seen most clearly in “modular” analyses of the mind. The human mind isargued to be modular in that vision and audition, face recognition, and thenumber sense are all separable faculties governed by their own generalizationsand principles. Chomsky’s work over several decades has provided a wealth ofevidence that the language faculty constitutes a separate module in this sense,akin in many respects to any other organ of the body (Chomsky 1975, 1984).Moreover, he has provided more, and more rigorous, evidence about the preciseinternal structure of this module than has been provided for any other domain,except perhaps vision. That is, there are two different notions of modularity:one according to which the language faculty is a module of the mind, distinctfrom moral judgment, music, and mathematics; another according to which the


Chomsky’s science of language 23

language module itself divides up into submodules, relating to sound, structure,and meaning. Evidence for both kinds of modularity comes from the indepen-dence one from the other of the various modules, as seen most clearly in doubledissociation.

Double dissociation

One can be blind without being deaf, deaf without being blind, and so on for allthe faculties with which we are endowed. Because our ears and eyes are separateorgans, the dissociation of deafness and blindness is unsurprising, even thoughthe cause of deafness or blindness may in some cases be due to damage to adifferent organ, the brain. That is, we expect the failure of a particular componentto lead to a particular deficit, even if, in the case of the brain, the complexityis so great that we may see a variety of different symptoms arising from thefailure of a single part. Moreover, the workings of these various components areindependent of intelligence, of the working of the “central system” in Fodor’s(1983) term. In the case of sight this is well understood: no one any longerexpects the misfortune of blindness to correlate with IQ, but the same appearsto be true of language.

A striking example of such dissociation is provided by the case of Christopher(Smith & Tsimpli 1995), a man who lives in sheltered accommodation becausehe cannot look after himself, who cannot solve problems of the intellectualcomplexity of noughts and crosses (tic-tac-toe), but who can nonetheless read,write, speak, and understand some twenty or so languages. An example of sub-modular dissociation within the language faculty can be seen in the fascinatingcase of MC (Froud 2000), who can read nouns and verbs of arbitrary complex-ity, but who cannot cope with “function” words like after, not, the, or because,at all.

Such cases show us that our knowledge of language is both more complex andmore isolable than might appear at first sight. It is time to look in some detail atprecisely what this knowledge consists of, as this is the area in which Chomskyhas made the greatest technical advances, and also the area which underpinsthe philosophical and psychological claims that have made him famous (orinfamous) in the rest of academe.

Knowledge of language

Our knowledge of English (or any other language) enables us to produce andunderstand any of an indefinitely large number of sentences and, additionally,we can make judgments of well-formedness about sentences we have neverpreviously encountered. Consider an example of the order of words possiblein simple sentences of English containing an adverb like occasionally. All the


24 Neil Smith

examples in (1) are equally acceptable but (2), despite the fact that it is perfectlyeasy to understand, is immediately perceived as odd:

(1) a. Occasionally John speaks French.b. John occasionally speaks French.c. John speaks French occasionally.

(2) John speaks occasionally French.

With other kinds of adverb, such as fluently, the range of possibilities is morelimited and the contrasts even starker. The example in (3) is fine, but those in(4) are of marginal acceptability and (5), like (2), is ungrammatical:

(3) John speaks English fluently.

(4) a. Fluently John speaks English.b. John fluently speaks English.

(5) John speaks fluently English.

If you are a linguist and your theory of word order entails that English speakersshould find (2) and (5) as acceptable as (1) and (3), then you are wrong, andyour analysis must be replaced by a better one. A corollary of this emphasis onseeking testable explanations is that the central concern of linguistics, as of theother natural sciences, is evidence rather than just data; this emphasis in turnentails that idealization is necessary: not everything can be considered.

In presenting all these examples I have been assuming that you will agree withmy judgments: that (1) and (3) are acceptable forms of speech, for instance.In fact, a major and innovative characteristic of Chomsky’s linguistics is itsexploitation of the previously neglected fact that we are able to recognize imme-diately that some sentences – like (5) – are ungrammatical: we have what onemight call “negative knowledge.” Hamlet can tell Ophelia that “I loved younot,” and we can understand him easily enough, but we know we have to say “Ididn’t love you”; similarly, Othello can ask Desdemona “Went he hence now?”,though we know we have to rephrase it in current English as “Did he go?” Suchknowledge is not restricted to the literary idiom. We can say equally well eitherof the sentences in (6):

(6) a. I asked the way to the school.b. I inquired the way to the school.

But whereas (7a) is fine, the analogous (7b) is odd, as indicated by the prefixedasterisk:

(7) a. I asked the number of people in the class.b. *I inquired the number of people in the class.



We should be as puzzled by the fact that we have these intuitions as by the factthat apples fall down not up, or that the sea has waves. Newton was not thefirst to notice apples falling, but his insight that why apples fall is in need of anexplanation led ultimately to his theory of gravity. Chomsky was not the first tonotice the elementary facts I have cited here, but his insight that our intuitions(our judgments) can tell us something profound about the human mind/brainis of comparable importance. As the biochemist Albert von Szent-Gyorgi putit: “Discovery consists of seeing what everybody has seen and thinking whatnobody has thought.”

I have appealed to various facts about our knowledge, but even decidingwhat the facts are is itself frequently unclear and dependent on other theoreticalcommitments. Not everyone agrees with the judgments given for (6) and (7),and it is not obvious whether the examples in (4) are to be assimilated interms of grammaticality to those in (3) or (5), or neither. If, like me, youfind them of intermediate status, there is an immediate and perhaps surprisingconsequence: it is not possible to group sentences exhaustively into those whichare grammatical and those which are ungrammatical, as there is a third categorywhich is neither one thing nor the other.

We turn now to take a closer look at what “knowledge of language” con-sists of, beginning with the fundamental distinction between “competence”and “performance.”

Competence, performance, and idealization

To be able to read this book you have to know English. It also helps if thelight is on. Turning the light out might stop you reading – your performance.It presumably wouldn’t affect your knowledge of English – your competence(Chomsky 1965). Slightly more subtle evidence for this dichotomy can begleaned from a variety of sources, but two kinds are particularly straightforward:the fact that our linguistic knowledge ranges over an infinite domain, and thefact that we can make mistakes. Consider first infinity, and take the propositionin (8):

(8) You should have pessimism of the intellect and optimism of the will.

The aphorism is due to the Italian thinker Antonio Gramsci, and has been quotedmore than once by Chomsky (1992b: 354) “No-one has ever said it better thanGramsci . . . You should have pessimism of the intellect and optimism of thewill.” As it encapsulates Chomsky’s attitude rather well, I have quoted his reportmyself (Smith 1999), and there is no linguistic reason why my students in turnshouldn’t quote me quoting Chomsky, quoting Gramsci . . . and so on, resultingin a sentence which is arbitrarily long. If there is no longest sentence, just asthere is no highest number, the set of sentences of your language is potentially


26 Neil Smith

infinite, even if the set you can actually get round to uttering is finite. There mayof course be non-linguistic reasons for not indulging in the repeated quotationsuggested here: the desire for originality, the fear of exhaustion, a finite life-span. But these are nothing to do with the grammar of English, or whatever lan-guage in which you choose to couch the report. Assuming – uncontroversially –that the mind/brain is finite, it must then be the case that there is some setof rules, a generative procedure4 for producing such sequences out of theirparts.5

Equally cogent evidence for the distinction between what we know and whatwe do with what we know comes from the phenomenon of mistakes, bothpathological and “normal.” Slips of the tongue are commonplace and may evenprovide evidence about linguistic structure (Fromkin 1973), but their interesthere is more elementary: the recognition that something is a mistake entailsthe existence of a norm from which it is a deviation; equivalently, that thereis a mismatch between competence and performance. Such mistakes occursporadically in all speakers, as a function of tiredness, carelessness, inebriation,and so on; in certain cases of pathology they may become all pervasive. Ifsomeone has a stroke and resulting (partial) loss of language, their speech maybe so replete with mistakes that they are hard or impossible to understand. Insome such cases the brain damage resulting from the lesion may have destroyedthe subject’s grammar; in others it may simply have rendered it inaccessible.In the latter case the patient may make a full recovery and regain the ability touse the language he or she appeared to have lost. As no one recovers from astroke able to speak a language different from any they knew before, it is safeto conclude that their knowledge of language was all the time intact, despitetheir temporary inability to exploit it.

Generative linguistics is about competence, but it should not be forgottenthat performance can provide crucial evidence for that competence. It is alsoimportant to emphasize that no kind of evidence is particularly favored in con-structing linguistic theories. It is sometimes thought that evidence derived frompsychological experimentation, or from brain scans,6 or from pathological casesof the sort referred to above, has some kind of priority or greater weight thanpurely “linguistic” evidence: i.e. evidence based on judgments of native speak-ers. This is a fallacy, as in general the latter is richer than the former, and allowsthe construction of theories with a deeper deductive structure.

If one is looking for evidence, certain facts are irrelevant, though it is a seriousproblem to decide which ones. In general, the role of scientific experimentationis to get us closer to the truth, to the ideal, by eliminating irrelevant extraneousconsiderations. In other words idealization reveals that which is real, but whichis usually hidden from view by a mass of detail. Most scientists accept the realityof the inverse-square law, whether it is being used to explain the intensity ofthe light reaching us from a star, of the sound reaching us from a jet engine,or the attractive force of a magnet, even though the messiness of experiments



means that their measurements never mirror it exactly, giving “a distortion ofreality” in Chomsky’s phrase (cited in Smith 1999: 12). Consider an exampleof idealization in the linguistic domain.

First-language acquisition takes place within a particular window ofopportunity known as the critical period (Smith 1998), which lasts for a fewyears and ends (at the latest) at puberty. Given what we know about this criticalperiod, it sometimes comes as a shock to read that first-language acquisitionis idealized to “instantaneity” (Chomsky 1986: 52). How can a process thatextends over several years be sensibly treated as though it took no time at all?The paradox is only apparent, not real. Although there is a striking uniformityacross children learning their first language in respect of the stages of develop-ment they go through, they do nonetheless differ from each other. For instance,in the course of mastering the system of negation, one child may form negativesentences by using an initial no, while another may use a final no, giving thecontrast in (9):

(9) a. No like cabbage.b. Like cabbage no.

Despite this developmental difference, both children will end up with the samesystem of negation in which they use the correct adult form in (10):

(10) I don’t like cabbage.

As far as we can tell, the early difference in the children’s systems has noeffect at all on the grammar they end up with. If the focus of our interest ison what’s known as “the logical problem of language acquisition” – children’stransition from apparently having no knowledge of a language to being able tospeak and understand it like adults – then we have support for the idealizationto instantaneity, which says that the different stages children go through in thelanguage-acquisition process are of no import to their ultimate psychologicalstate. Of course, it may turn out that this surprising claim is false. It might be thatsufficiently sophisticated tests of grammaticality judgment, or investigationsof neural firing, or subsequent historical changes in the language concerned,showed that the two children’s grammars were crucially different, and differentin ways that could explain other mysterious facts about knowledge of language.It’s possible, but there is (as yet) no evidence, and the idealization is accordinglyjustified: it leads us to an understanding of one of the variables in the real systemwe are studying.

Levels of representation

A large part of the Chomskyan enterprise has been devoted to making explicitexactly what our “knowledge of language” consists of. To find out, it is necessaryto look at the notion “level of representation.”


28 Neil Smith

As we saw in the discussion of modularity, the language faculty can beviewed from the inside or from the outside. From the outside it has to provide aninterface with other components of the mind/brain; from the inside it has to relatemeanings to pronunciations, semantics to phonetics. The formal properties ofthe mechanisms effecting this latter relation are to a considerable extent areflection of the need to make the products of the language faculty “legible”to these other systems. In particular, the complexity of the inner workings ofthe grammar can be seen on analysis to be largely the side-effects of the needto interact with non-linguistic systems, and the linguistic system itself is assimple as is possible: grammars of natural languages are a “perfect” solutionto the problem of relating sound to meaning (Chomsky 2000a: 9f). We returnto this striking suggestion after we have looked at the internal structure of thegrammar.

The first distinction that needs to be made is that between the lexicon and the“Computational System,” basically the difference between what you have tostore in memory and what you can create anew as the occasion demands. Youhave to remember that cat means a certain kind of animal, but you do not haveto remember – and in principle could not remember – the potentially infiniteset of sentences about cats that you are capable of producing or understanding.

From the earliest work in generative grammar (Chomsky 1951, 1955, 1957a)the notion “level of representation” has been central, where different levels ofrepresentation are postulated to capture generalizations of different kinds aboutsentences. To express regularities about pronunciation and the sound structureof sentences, the grammar exploits the level of representation called PhoneticForm (PF). To capture generalizations about meaning and the logical propertiesof sentences, it exploits the level of semantic representation called Logical Form(LF). The existence of ambiguous sentences like (11):

(11) Picasso painted his models nude.

means that the relation between the LF and the PF is many to one: such sentenceswill have the same phonetic form but different logical forms. Similarly, distinctsentences that mean the same thing, like those in (12):

(12) a. All the children came.b. The children all came.

entail that the relation between PF and LF is likewise many to one: such sen-tences will have the same “logical form” but different “phonetic forms.” Theimplication of this many–many relation is that the mapping from sound to mean-ing is indirect, mediated by the syntactic structure of the sentences involved. Inthe same way, the partial similarity (and partial difference) between quantifierssuch as every and each is described by assuming distinct semantic specifications



for them in the lexicon. This accounts directly for the semantic contrast in (13)and indirectly for the contrast in grammaticality in (14):

(13) a. Each child came.b. Every child came.

(14) a. The children each came.b. *The children every came.

In order to give an adequate description of these semantically conditioned dif-ferences of syntactic behavior, it is necessary to have a level of LF at which theprecise meaning of every expression is specified.

The complexity of syntax

Apart from variations in technical terminology, so much has always been com-monplace. What was striking about Chomsky’s work from the beginning was thesophistication – and complexity – of his syntactic representations. In particular,he argued at length for the necessity of having more than one level of repre-sentation within the syntax, postulating the famous distinction between “deepstructure” and “surface structure.” This terminological contrast didn’t featureexplicitly in the earliest work, but it is implicit both there and in any frame-work which exploits the notion of “transformational derivation.”7 In Chomsky’scurrent (minimalist) framework (Chomsky 1995c; Uriagereka 1998), neither ofthese syntactic levels is actually necessary, but seeing why involves first lookingat the various rule types that are used in a grammar. It is, however, worth not-ing now that a minimalist approach is essentially just scientific common sense.Chomsky’s claim that linguistic theory should postulate only those notions (e.g.levels of representation) that are “either conceptually necessary or empiricallyunavoidable” is taken by scientists as so obvious as not to need saying.

Rules

We have already seen that rules are necessary to characterize the infinite pos-sibilities provided by natural language, but they have a number of other usefulproperties. Levels of representation are defined by the rules that yield them astheir output. Rules let us capture significant generalizations that go beyond lex-ical idiosyncrasy (all verbs have a past tense, not just the verb evolve); and theymake possible an account of semantic compositionality – the fact that the mean-ings of larger entities is made up of the meanings of smaller ones: the meaningof I don’t like cabbage is built up out of the meanings of I, not, like, and cabbage.It follows from these properties that sentences are not just strings of words buthave structure. This is most obvious when the same sequence of words has


30 Neil Smith

two different interpretations depending on the relations between them, as in theambiguous black cab driver which can mean either “a black driver of cabs” or“a driver of black cabs.” The difference can be shown by using brackets to setoff the constituents out of which sentences are constructed: in the present case,as shown in (15):

(15) a. [black [cab driver]]b. [black cab [driver]]

Linking two or more words together as a constituent is not an innocent exercise:it predicts that the same combination will recur, explicitly or implicitly, in otherexpressions. In (16a), the sequence many people is a constituent which showsup in a different position in (16b):

(16) a. Many people are in the room.b. There are many people in the room.

Separating the two elements (many and people) of this constituent gives rise(in this case) to ungrammatical sequences of the kind in (17):

(17) a. *Many there are people in the room.b. *How many are there people in the room?

Similarly, a constituent can typically be replaced by a “pro-form,” such as they,in (18a), or an empty category as in (18b):

(18) a. There are many people in the room but they are hiding.b. Many people are in the room but – are hiding.

where the dash shows the position of the “understood” empty subject of hiding.To account for facts of this kind Chomsky and linguists working within his

paradigm postulate two different kinds of rule: one that “merges” two words toform a larger constituent: e.g. many and people; and another that may “move”such a constituent to a different position: e.g. moving many people (but notjust many) to the front of the sentence.8 Such rules operate under well-definedconditions: the relation between many people and they, or between many peopleand the empty –, can obtain only in certain circumstances.9 In (19a) the itemthey may (but need not) be taken to refer to many people, but in (19b) this isimpossible:

(19) a. Many people think they are intelligent.b. They think many people are intelligent.

In fact, we clearly know far more than this even about such simple sentences asthose given. Corresponding to the statements in (16), we have such questionsas those in (20):



(20) a. Are many people in the room?b. Are there many people in the room?

The first can be answered appropriately as in (21):

(21) Many people are in the room and many people aren’t in the room

(where, for instance, half are indoors and half are in the garden). A comparableanswer to the second, as in (22), is anomalous:

(22) There are many people in the room and there aren’t many people inthe room.

We still have the same constituents – many people, in the room, and so on, butthe interpretations are different.

All of us have knowledge of this kind, even though most of us are not aware ofit, and none of us can spell out this knowledge in full detail. The notion of suchunconscious or “tacit” knowledge has been anathema to many philosophers(Quine 1972), but the issue has been bedevilled by often arbitrary stipulationsas to what constitutes “knowledge.” The important point is that humans canproduce and understand any of an infinite number of largely novel sentences,and can make systematic judgments about their well-formedness, and we needto explain these abilities. The explanation in terms of merge and move may bewrong, but it is currently both the most inclusive and the deepest in terms ofthe deductive structure of the theory involved.

Displacement

The possibility of movement of items is invoked to account for what isoften called the “displacement property” of language (Chomsky 2000a: 12f):constituents are pronounced in places other than where they are interpreted.The simplest kinds of example in English are provided by the contrast betweenstatements on the one hand, and questions and focus constructions of the kindseen in (23) on the other:

(23) a. These delegates might elect the best candidate.b. Which candidate might these delegates elect?c. The best candidate, these delegates might elect

The basic intuition about such examples is that the phrase containing candidateis in all cases the object of elect, even though it appears either at the beginningof the sentence or after the verb. Elect is a two-place predicate: it needs a subjectand an object, respectively these delegates and the best candidate in (23a), sothat the failure of either to appear results in ungrammaticality: all the strings in(24) are unacceptable:


32 Neil Smith

(24) a. *these delegates electedb. *elected the best candidatec. *elected

To account for this, elect is said to select an object to its right, and this objectmay then remain where it is, or be displaced, as in (23b).10 This selectionaldifference simultaneously accounts for the contrast with, on the one hand, verbslike giggle, which allow no object, as witness the impossible (25), or, on theother hand, verbs like divide which can be used with or without an object, asin (26):

(25) Don’t giggle me.

(26) a. The amoeba divided.b. They divided the spoils.

In fact (25) is sometimes produced by children (Bowerman 1987), though it isnot likely to be acceptable to anyone reading this – another instance not onlyof our “negative knowledge,” but of our intriguing ability – in the absence ofany obvious evidence – to progress from a grammar which allows (25) to onewhich does not. This is a classic example of poverty of the stimulus: we knowmore than the environment provides evidence for. It is implausible to suggestthat every child who comes up with examples like (25) is explicitly correctedby parents or peers, yet the child pattern is common and the adult intuition isrobust.

Interestingly, all languages allow for statements and questions of the kind in(23), but in some languages (such as Chinese and Japanese), the object stays inthe same place and doesn’t get displaced (Chomsky 1995c: 69), giving rise tosentences like (27):

(27) These delegates might elect which candidate?

which is itself acceptable as an echo-question in English: that is, a questionrepeating with incredulity something already suggested. We thus have a sit-uation in which there are two different syntactic structures in two differentlanguages, but with the same interpretation. That is, at some level of represen-tation – presumably LF – both (23b) and the Chinese equivalent of (27) havethe same logical structure.

That is, just as we have overt and covert syntactic categories (they versusnothing in (18)), we have the possibility of overt (visible) and covert or invisiblemovement of some categories. The parallel interpretations of (23b) and (27)are captured by saying that in English the movement seen in (23b), giving thesyntactic and logical representation, is “overt,” whereas the same displacementin Chinese etc. is “covert.” In the jargon, these are referred to as before and after“Spell-out”: movement before spell-out is visible – as in English; movement



after spell-out is invisible – as in Chinese. That the movement is comparable inthe two cases is evidenced by a wealth of syntactic facts as well as the intuitivelyobvious identity of interpretation, where such identity is taken to entail identityof representation at some level.

Empty categories

The idea that there should be phonologically empty words of the kind men-tioned here has either outraged or bewildered many members of the linguisticcommunity. Yet the development of a theory of empty categories has beenimportant in allowing linguists to capture interesting generalizations, to sim-plify the structure of grammars, and ultimately to provide evidence about theinnate endowment that the child brings to the task of first-language acquisition.

Traditional descriptions of language frequently refer to “understood” ele-ments, which are not visible (or audible), but whose presence it is convenient toassume. This tradition was adopted and formalized in the transformational treat-ment of the subject of coordinate sentences like (18b), as well as for imperativesand a variety of other constructions. More recently, it has been widely extended,with interesting implications for the structure of the grammar. Consider the pairof sentences in (28):

(28) a. John wants Bill to go.b. John wants to go.

It is intuitively obvious that in (28a) Bill is to go and in (28b) that John is togo. Making that intuition grammatically explicit can be effected by assumingthat in each case go has a subject, even though that subject is invisible in (28b),whose structure is then something like (29):

(29) John wants [ec] to go.

where “ec” stands for an “empty category,” construed as referring to the sameperson that John does, as is explicit in the synonymous John wants himself togo.

An empty category is in general one that has syntactic properties but is notpronounced. With sentences as simple as these the gain from assuming emptycategories is scarcely overwhelming, but in more complex cases, one can beginto see how benefits accrue. For me, and perhaps most speakers of English,a sentence like (30a) typically allows the alternative pronunciation shown in(30b), where I am contracts to I’m:

(30) a. I am the greatest.b. I’m the greatest.


34 Neil Smith

Stating the precise conditions under which such contraction is possible is notstraightforward – (31b) is simply ungrammatical:

(31) a. John is planning to come at the same time as I am.b. *John is planning to come at the same time as I’m.

(31a) is interpreted as meaning “John is planning to come at the same time asI am planning to come.” The italicized words are redundant and so are typi-cally not pronounced, even though they are “understood,” and have a syntacticexistence in the form of an empty category. This suggests an explanation forthe impossibility of (31b): contraction cannot take place adjacent to an emptycategory; am is adjacent to an empty category, left by the omission of planningto come, so (31b) is excluded (Lightfoot, this volume).

Another initially mysterious contrast, illustrated in (32), succumbs to thesame explanation:

(32) a. Tell me whether the party’s tomorrow.b. Tell me where the party is tomorrow.c. *Tell me where the party’s tomorrow.

Why is (32c) ungrammatical? An answer is suggested by a consideration of thekind of echo-question seen in (27), They elected which candidate?, and furtherillustrated in (33):

(33) a. The party’s where tomorrow?b. The party’s when tomorrow?

In these examples where and when appear in the same position as the ordinarylocative or temporal phrases that they question, as seen in (34):

(34) a. The party’s in the hangar tomorrow.b. The party’s at 11 o’clock tomorrow.

The next stage of the argument should be clear – the structure of (32b) is asshown in (35):

(35) Tell me where the party is [ec] tomorrow

with an empty category marking the place from which where has moved, andblocking the contraction of is to ’s.



I-language and E-language

It is time to take stock. We uncontroversially have knowledge of language,manifest in our ability to produce and understand any of indefinitely manysentences, in our ability to make judgments of well-formedness and, moreimportantly, of ill-formedness: we can recognize previously unheard mistakesor deviations from the patterns of our native language. It is important to notethat the strength of the argument from the existence of mistakes does not rest onthe identification of a common language shared by the speaker and the hearer.If you hear someone speaking a different dialect of your language you do notassume that they are making mistakes in pronouncing words slightly differently,or in using different vocabulary or grammar. For me, it’s not a mistake to saypavement rather than sidewalk, or to make sort rhyme with bought: that’s justthe way I speak my variety of (British) English. Similarly, I use the forms in(36) whilst others use those in (37) to convey the same message:

(36) a. I jumped out of the window.b. I jumped off the table.

(37) a. I jumped out the window.b. I jumped off of the table.

But this raises a basic issue that differentiates Chomsky’s linguistics from thatof all his predecessors: his concentration on the language of the individual ratherthan the language of the social or political group in which that individual resides.For much of the time the difference doesn’t matter, but there are importantphilosophical implications of the difference, encapsulated in the terminologicaldifference between “I-language” and “E-language” (Chomsky 1986).

When generative grammar was being first developed (Chomsky 1955, 1957a),a language was defined as a set of sentences, generated by the rules of a grammar.Example (3) (John speaks English fluently) would be generated, and so be partof the language; (5) (John speaks fluently English) would not be generated,and so would not be. Chomsky’s early work included a demonstration that anysuch definition of “language” could not have a decisive role to play in linguistictheory. The idea that the E-language, some definable entity external to theindividual and that corresponds to the everyday idea of “English,” could be thefocus of theory construction is not tenable. If such a social or supra-personalconstruct was coherent and consistent, it might be the appropriate domain forpolitical, mathematical, or logical statements, but if it is supposed to reflect thehuman capacity for language, it is neither coherent nor definable.11 By contrast,I-language, an individual’s internal possession, with its explicitly psychologicalstatus, is the appropriate domain for statements about personal knowledge, andis also coherently definable. It is uncontroversial that untutored speakers can


36 Neil Smith

make grammatical judgments of the kind I have been appealing to throughoutthis discussion: someone who failed to be sensitive to a difference between(3) and (5) would simply not know English. Chomskyan linguistics aims tocharacterize explicitly what underlies that ability. In contrast, it is not at allobvious that there is a coherent domain of theoretical investigation in whichthe geopolitical notion of “English” should figure. For me, both the sentencesin (38) are possible:

(38) a. Nothing, I had for breakfast.b. Nothing did I have for breakfast.

For some of my colleagues only one of them is acceptable (Cormack & Smith2000). It is an interesting research question to tease out the differences betweenour respective grammars; but it doesn’t make much sense to ask which of usis correct, or which is “English.” In brief, the apparently narrower domain ofI-language – the mentally represented grammar of an individual – is amenableto scientific investigation in a way that E-language is not. The subject matterof linguistics is (tacit) knowledge, and no single individual has or could haveknowledge of all the varieties of any language in the traditional lay sense of theterm.

As we have seen, linguists attempt to formulate aspects of our knowledgeof language in the form of rules that both make explicit what we know andthereby make predictions about what else can occur in our own languagesand in human language more generally. In principle, such predictions makefalsifiable claims, but it is important to note that no serious theory adopts thekind of naıve falsificationism supposedly proposed by Popper (1963). Individualanalyses of particular data will be refined or rejected in part on the basis oftheir conformity with such predictions, but the guiding intuitions behind thetheory can often be maintained despite apparent falsification. This apparentlycavalier attitude is a direct function of the idealization inherent in producingany scientific explanation. In linguistics as in physics, what is important is toexplain some subset of the data, rather than describe everything.

Legibility

A theory of language provides a link between sound and meaning, betweenrepresentations of the pronunciation and representations of the logical propertiesof words and sentences. Accordingly, a grammar – the I-language – must definethe two levels of representation, PF and LF, and specify the link between them.Ideally, there should be no other levels and the complexity of this link shouldbe minimal. This suggests two questions which it had previously either beenimpossible to address seriously or perhaps even to formulate. First, how gooda solution to this conceptual problem of linking sound and meaning is a human



language? Is it right to suggest that the grammars of natural languages arein some sense optimal? Second, what are the relations between the languagefaculty and other systems of the mind/brain? In particular, can any perceiveddeviations from optimality in the first be attributed to conditions imposed bythe second?

Chomsky addresses these issues in terms of the question: “how ‘perfect’ islanguage?”, with the answer, surprising for a biological system (2000a: 9),12

that it is very close to perfect. What this means is that any deviations fromconceptual necessity manifested by the language faculty (that is, the I-language)are motivated by conditions imposed from the outside. Chomsky calls these“legibility conditions”: conditions imposed by the need for other systems ofthe mind/brain to use representations provided by the language faculty. Thisrefers in particular to the need for the articulatory and perceptual systems toexploit PF representations, and for the conceptual system to exploit LF. Againstsuch a background, movement or “displacement” processes of the kind seen in(23) or (38), or in the different positions occupied by Clinton in They electedClinton and Clinton was elected, appear to be conceptually unnecessary. Whydo natural languages exploit such devices which are completely foreign to theartificial languages of logic and mathematics? One tentative answer is thatdisplacement may plausibly be motivated by the need to structure informationfor optimal communication. To elaborate a little: the “focus” example in (23c)is probably motivated by the desire to make more salient one particular entityin the discourse by putting it in initial position. The subjects one wishes to talkabout are likely to be differentially accessible to one’s interlocutor, dependingon how recently they have been mentioned, how prominent they are in theenvironment, and so on. Putting a constituent like the best candidate in initialposition then increases its accessibility, and makes fruitful communication morelikely. If this is indeed the correct account then it looks as if a property of thelanguage faculty is imposed from outside the system, from another part of themind/brain: from the “central system” in Fodor’s (1983) terminology.13

Chomsky does not stop there, but attempts to link this apparent imperfec-tion of language to another. Natural languages are full of phenomena that giverise to problems for second-language learners and irritation for philosophers.There are morphological complexities like declensional paradigms and irregularverbs, which appear to have no real meaning of their own and to be semanticallyuseless. They are another imperfection, necessitating the postulation of unin-terpretable features – that is, features with no semantic interpretation. However,current syntactic theory makes systematic use of such uninterpretable features:their function is to drive the movement processes that we have just seen to bemotivated from outside the language faculty. If such conjectures are on the rightlines, they allow the interesting possibility of reducing two kinds of apparentimperfection to one. In fact, if the argument is correct, the imperfections are


38 Neil Smith

indeed only apparent. Given the constraints that other systems of the mind/brainimpose on solutions to linking sound and meaning, there may be no other alter-natives, so conceptual necessity explains the form of the grammar overall.

Although we may differ in our ability to deploy our knowledge of language,that knowledge is largely the same from person to person. As far as is known,we all see the same way: across the species we have the same visual system;similarly, across the species, we all have the same linguistic system. The caseof language is somewhat different from that of other putative modules in thatit is evident from the existence of languages other than our own that a certainamount of language is acquired on the basis of interaction with the environment:it is “learnt.” However, the most fundamental aspects of language are universaland can hence be factored out of the learning equation. In these areas we do notexpect there to be differences between individuals, whether they are speakersof different languages or the same language. To give a simple example: we donot need to learn that our language contains nouns and verbs: all languagescontain nouns and verbs. What we need to learn is which noises our languageassociates with particular examples of them.

Acquisition

Principles and parameters

So some of the intricate and complex knowledge characteristic of adult grammaris putatively innate, but some of it is obviously “learnt.” More accurately, it isattributable in part to the role of the ambient environment: languages differalong various dimensions, and it is obvious that we are not born knowingthat the words tortoise and turtle pick out chelonians, or that English puts thesubject before the verb, whereas Welsh puts it after. In recent years, the ideathat linguistic theory should be “explanatory” has been focused on explaininghow our knowledge of language can be acquired, given the (claimed) povertyof the stimulus.14

In the last twenty years, for the first time in the history of language study,there is now a reasonable chance of solving “Plato’s Problem” – how it is wecan know so much given that the evidence is so impoverished.15 The solution isPrinciples and Parameters. The idea is that the child is born knowing the princi-ples which determine the so-called design features of language: the existence ofparticular categories such as Noun and Verb, the possibility of Merge and Move,the universality of structure dependence, the conceptual structure of possiblepredicates like cause and location (Bloom 2000), and much more.

A central aspect of Principles and Parameters theory is that it explicitlyexploits the idea that there is a cascade effect so that knowledge can developwithout being learned. There is evidence that children acquiring their first



language home in on the correct word order extremely early. Before they canspeak themselves, children have worked out that English places the verb beforeits object, so we have examples like eat this rather than this eat. In the jargonthis is usually referred to as being a reflex of the fact that English is “head-first”:the head verb eat precedes its complement this. Once this elementary fact hasbeen established, children also know a number of other facts about English:that we say fond of tortoises rather than of tortoises fond, the idea that penguinsare cute and not that penguins are cute’s idea. They do not need to have heardsuch examples to have the knowledge. They are born with the principle andwith some specification of the range of variation in possible human languages,and on being presented with data they need merely to pick out which of thesepossible languages is the one that they are being exposed to. The process is oneof selection (Chomsky 1980; Piattelli-Palmarini 1989) from an antecedentlydefined set, rather than instruction about an unconstrained system. One kind ofevidence for this theory comes from the interesting category of “mistakes thatchildren do not make.” Over-generalizations of the kind three sheeps comed arepervasive in the morphology, but comparable analogies typically do not occurin the syntax. Observing that in a sentence like (19a) (Many people think theyare intelligent) the pronoun they may or may not be dependent on many people,the child typically does not extend the possibility to (19b) (They think manypeople are intelligent) where the they may not be dependent on many people.That these mistakes just don’t appear suggests that children are predisposed tointerpret the input they are exposed to in particular ways: the principles andparameters are there from the beginning.

Typology

The theory of parametric variation identifies precisely those aspects of language(apart from the morphophonological makeup of individual items in the lexicon)which it is necessary for the child to acquire on the basis of experience. Thetheory then has as a natural corollary that these aspects are precisely those byreference to which languages can differ one from the other, and hence the the-ory simultaneously provides an account of possible typological variation amonglanguages. In each case, the locus of this variation is the set of functional cat-egories – categories like Determiners, Tense and Complementizer. This is notsuperficially obvious, but consider again the contrast between (23b) and (27),where in different languages WH words (who, which, etc.) may either moveto the front of the sentence or stay where they are. In the case of English andsimilar languages where the WH phrase moves, there is a technical question asto where it moves to and why it moves at all. The answer to the first question isthat it moves to a position called the “Specifier” (Spec) of the ComplementizerPhrase (CP), the answer to the second is that it does so because (in English but


40 Neil Smith

not in Chinese) the C is “strong” and acts like a magnet attracting the item to it.The technical details are spelt out in the relevant literature (Uriagereka 1998).Suffice it here to say that all sentences have a CP, but that this position maybe either empty or filled. It is typically, though optionally, filled in subordinate“complement” clauses as in (39), where the complementizer that or the com-plementizer phrase which candidate is shown in italics, and the parenthesesindicate optionality:

(39) a. I know (that) they elected the best candidate.b. I know which candidate they elected.

That the two items are in the same position is indicated by the impossibility ofhaving both of them together, as shown by the unacceptability of (40):

(40) *I know that which candidate they elected.

On the most economical assumptions about the rules of the grammar, it mustbe that in (23b) which candidate has moved to the same position as in (39b),hence both must share the same landing site for the “moved” element: viz.Spec CP. From this it follows that quite radical differences between languagescan be accounted for in terms of properties of functional categories, despitethe involvement of substantive elements like nouns and noun phrases. Moreparticularly, the differences are limited to the presence of specific features onfunctional categories within the lexicon. The lay intuition that the basic differ-ences between languages reside in their vocabulary has been given substanceby being recast in an explicit theoretical framework.

Universals and evolution

If a substantial part of our knowledge of language is innate, and if any childcan learn the language it is immersed in with equal facility, it follows that manyaspects of our knowledge are universal. So much is now generally accepted(with some disagreement about the details of what is innate). It is nonethelessstill something of a shock for many to read that one can gain insight into Englishby studying Japanese (Chomsky 2000a: 53–4), or any other language; and ofcourse vice versa. The logic is clear: suppose that a phenomenon in languageA can be adequately described by using either of two devices – d1 or d2 – butthe comparable phenomenon in language B can only be adequately describedby one of them, say d1, then theoretical parsimony demands that one use d1 forboth languages, rather than d1 for one and d2 for the other. On further analysisit should also turn out that this choice makes appropriate further predictions,but not always: the poverty of the inflectional system (of English, for instance)may make it impossible to test some hypotheses in that language, whereas theycan be easily tested in another (Cole & Hermon 1981; Smith 1999).



We have seen that knowledge of language is rich and intricate, but is in somesense perfect. Moreover, it is largely innate and common to the species, andhence is presumably genetically determined. This rather surprising combinationof properties is explicable perhaps only if one can mount an explanation forthem in terms of evolution. If our ability to develop language is geneticallydetermined, but our relatives – from chimpanzees to mushrooms – have nocomparable language faculty, then it must have evolved. But now there is aproblem, as Chomsky is widely quoted as denying that natural selection couldhave produced human language. Furthermore, as there are supposed to be onlytwo possible explanations for evolution, namely “God or natural selection,”Chomsky must either be appealing to divine providence or be a mystic. As hasrecently been documented in great detail (Jenkins 2000), Chomsky’s positionis not that natural selection has played no role in the evolution of language, butthat natural selection is only one factor. In all evolution, whether of language,the eye, or an antibody, an essential contribution is also made by physical anddevelopmental factors working on the variation provided by random mutation.The criticism of Chomsky with regard to evolution is based on a simplisticanalysis of the range of possibilities available, and of his position on them.Properties of language have evolved. Natural selection must have played a rolein this evolution, but so too have elementary physical constraints, such as the sizeof the human head. Many other factors have also, presumably, been involved,such as the adoption of a trait in one domain for use in another: an examplemight be the exploitation of discrete infinity by both the number sense andthe language faculty. Our knowledge is vanishingly small in this domain, butthe questions are coherent, the issues are empirical, and the problem of unifyingthe various bits of our knowledge with the rest of the natural sciences is standard.

It is appropriate to end on a note of unification. Linguistics cannot be reducedto any other science, but its findings can be unified with those of psychology,biology, and ultimately neurology. The lead in this unification is provided bylinguistics, because its theories are the most fully developed. Overwhelmingly,this is due to Chomsky.


2 Plato’s Problem, UG, and the language organ

David Lightfoot

The empiricist view is so deep-seated in our way of looking at the human mindthat it almost has the character of a superstition.

Chomsky (The Listener May 30, 1968)

Plato’s Problem

Plato’s Problem was expressed generally by Bertrand Russell: “How comesit that human beings, whose contacts with the world are brief and personaland limited, are nevertheless able to know as much as they do know?” Theproblem arises in the domain of language acquisition in that children attaininfinitely more than they experience. Literally so, we shall see: they attain aproductive system, a grammar, on the basis of very little experience. So thereis more, much more, to language acquisition than mimicking what we hear inchildhood, and there is more to it than the simple transmission of a set of wordsand sentences from one generation of speakers to the next. There is more toit than a reproduction of experience and, in maturity, our capacity goes wellbeyond what we have experienced.

Consider some subtleties that people are not consciously aware of. The verbis may be used in its full form or its reduced form: people say Kim is happy orKim’s happy. However, certain instances of is never reduce, for example, theunderlined items in Kim is happier than Tim is or I wonder what the problem is inWashington. Most people are not aware of this, but we all know subconsciouslynot to use the reduced form here. How did we come to this? The questionarises because the eventual knowledge is richer than relevant experience. Aschildren, we heard instances of the full form and the reduced form, but we werenot instructed to avoid the reduced form in certain places; we had no accessto “negative data,” information about what does not occur. Yet, all childrentypically attain the ability to use the forms in the adult fashion, and the abilityis quite independent of intelligence level or educational background. Childrenattain this ability early in their linguistic development. More significantly, child-ren do not try out the non-occurring forms as if testing a hypothesis, in the

42


Plato’s Problem, UG, and the language organ 43

way that they “experiment” by using forms like goed and taked. The abilityemerges perfectly and as if by magic. And it emerges despite limited relevantexperience, in a way that might have intrigued Plato.

Another example. Pronouns like she, her, he, him, his sometimes may referback to a noun previously mentioned in a sentence (1a–c). However, one canonly understand (1d) as referring to two men, Jay and somebody else; here thepronoun may not refer to Jay, unlike in (1a–c).

(1) a. Jay hurt his nose.b. Jay’s brother hurt him.c. Jay said he hurt Ray.d. Jay hurt him.

As adults, we generalize that a pronoun may refer to a preceding noun exceptunder very precise conditions (1d). But then, how did we all acquire the rightgeneralization, particularly knowledge of the exception?

Recall the nature of our childhood experience: we were exposed to a hap-hazard set of linguistic expressions. We heard various sentences containingpronouns; sometimes the pronoun referred to another noun in the same sen-tence, sometimes to a person not mentioned there. Problem: because we werenot informed about what cannot occur, our childhood experience provided noevidence for the “except” clause, that pronouns sometimes do not corefer. Thatis, we had evidence for generalizations like “is may be pronounced z” and“pronouns may refer to a preceding noun,” but no evidence for where thesegeneralizations break down.

As children, we came to know the generalizations and their exceptions, andwe came to this knowledge quickly and uniformly. Yet our linguistic experiencewas not rich enough to determine the limits to the generalizations. We call thisthe problem of the poverty of the stimulus. This is “Plato’s Problem” and ithas shaped much of grammatical theory in the work of Chomsky and others.Children have no data that show them that is may not be reduced in somecontexts and they have no data showing that him may not refer to Jay in (1d).These two small illustrations are examples of the form that the poverty-of-stimulus problem takes in language.

There are two “easy solutions” to the poverty-of-stimulus problem; neitheris adequate. One is to say that children do not overgeneralize, because they arereliable imitators. That is, children do not produce the reduced is in the wrongplace or use a pronoun in (1d) wrongly to refer to Jay, because they never hearlanguage being used in this way. In other words, children acquire their nativelanguage simply by imitating the speech of their elders. We know this approachis not tenable, because everybody constantly says things that they have neverheard. We express thoughts with no conscious or subconscious considerationof whether we are imitating somebody else’s use of language. This is true of


44 David Lightfoot

the most trivial speech: in saying I wanna get a ticket for the game that’s hereon Wednesday, one is using a sentence that one has almost certainly not heard.

Sometimes it is said that children form new sentences “by analogy” withwhat they have heard, but this simply conceals the problem: it does not accountfor why some analogies are drawn and others not. The problem is to explainwhy a reduced ’s in Tim’s happy provides an analogical basis for contracting thefirst is in Kim is happier than Tim is, but not the second. Why do the sentences(1a–c) not provide an analogical basis for coreference between Jay and him in(1d)? Children converge on specific generalizations and not on other logicallypossible ones, in ways that cannot be explained by a general notion of inductionor analogy.

A variant on this approach is that children learn not to say the deviant formsbecause they are corrected by their elders. Some aspects of language are taughtin schools: spelling conventions, socially stigmatized forms, some kinds oftechnical vocabulary. However, language emerges without the aid of teaching;many people are illiterate but nonetheless have a productive capacity to uselanguage. Furthermore, the appeal to instruction does not explain languageacquisition. First, it would take an acute observer to detect and correct theerror. Second, where linguistic correction is offered, young children are highlyresistant and just don’t get the correction. Third, in the examples discussed,children do not overgeneralize and therefore parents have nothing to correct;this will become clearer when we discuss experimental work on young children.

So the first “easy” solution to the poverty-of-stimulus problem is to deny thatit exists, to hold that the environment is rich enough to provide evidence forwhere the generalizations break down. The problem is real and this “solution”does not address the problem.

The second “easy answer” also denies that there is a problem, but it denies thatthere is anything to be learned, and holds that a person’s language is fully deter-mined by genetic properties. Yet this answer also cannot be right, because peo-ple speak differently, and many of the differences are environmentally induced.There is nothing about a person’s genetic inheritance that makes her a speakerof English; if she had been raised in a Dutch home, she would have become aspeaker of Dutch.

The two “easy answers” either attribute everything to the environment oreverything to the genetic inheritance. Neither position is tenable. Instead, lan-guage emerges through an interaction between our genetic inheritance and thelinguistic environment to which we happen to be exposed. English-speakingchildren learn from their environment that the verb is may be pronounced izor z, and native principles prevent the reduced form from occurring in thewrong places. Likewise, children learn from their environment that he, his,etc. are pronouns, while native principles entail where pronouns may not refer



to a preceding noun. The interaction of the environmental information and thenative principles accounts for how the relevant properties emerge in an English-speaking child.

I’ll sketch the relevant principles in a moment. A modern Plato might teasenative principles apart from learned elements and claim that the native prin-ciples reflect knowledge attained in a previous life, that the knowledge wasrendered subconscious when the soul drank from the River Lethe, the river offorgetting, just before birth. However, Chomsky assumes that native principlesare encoded somehow in genetic material. This involves a kind of Mendeliangenetics. In the mid nineteenth century, Mendel postulated genetic “factors”to explain the variable characteristics of his pea plants, without the slightestidea of how these factors might be biologically instantiated. Similarly, linguistsseek to identify information which must be available independently of experi-ence, in order for a grammar to emerge in a child. We have no idea whetherthis information is encoded directly in the genome or whether it results fromepigenetic, developmental properties of the organism; it is, in any case, native.As a shorthand device for these native properties, I shall write of the linguisticgenotype, that part of our genetic endowment which is relevant for our linguis-tic development. Each individual’s genotype determines the potential rangeof functional adaptations to the environment (Dobzhansky 1970: 36), and weassume that the linguistic genotype (what linguists call “Universal Grammar”or “UG”) is uniform across the species (short of pathological cases). That is,linguistically we all have the same potential for functional adaptations andany of us may grow up to be a speaker of Catalan or Hungarian, depend-ing entirely on our circumstances and not at all on variation in our geneticmake-up.

Since children are capable of acquiring any language to which they happento be exposed between infancy and puberty, the same set of genetic prin-ciples which account for the emergence of English (using “genetic” now inthe extended sense indicated) must also account for the emergence of Dutch,Vietnamese, Hopi, or any other of the thousands of languages spoken by humanbeings. This plasticity imposes a strong empirical demand on hypotheses aboutthe linguistic genotype; the principles postulated must be open enough toaccount for the variation among the world’s languages. The fact that peopledevelop different linguistic capacities depending on whether they are broughtup in Togo, Tokyo, or Toronto provides a delicate tool to refine claims aboutthe nature of the native component.

Chomsky’s approach to Plato’s Problem, outlined in the first chapter ofAspects of the Theory of Syntax (1965), is to say that there is a biologicalentity, a finite mental organ, which develops in children along one of a numberof paths. The paths are determined in advance of any childhood experience. The


46 David Lightfoot

language organ that emerges, the grammar, is represented in the brain and playsa central role in the person’s use of language. We have gained some insightinto the nature of people’s language organs by considering a wide range ofphenomena: the developmental stages that young children go through, the waylanguage breaks down in the event of brain damage, the manner in which peo-ple analyze incoming speech signals, and more. At the center is the biologicalnotion of an internal, individual language organ, a grammar.

The nature of grammars

Children acquire a productive system, a grammar, in accordance with therequirements of the genotype. If asked to say quite generally what is nowknown about the linguistic genotype, one might say that it permits finite gram-mars, because they are represented in the finite space of the brain, but that theyrange over an infinity of possible sentences. Finite grammars consist of a set ofoperations which allow for infinite variation in the expressions which are gen-erated. The genotype is plastic, consistent with speaking Japanese or Quechua.It is modular, and uniquely computational.

By “modular” I mean that the genotype consists of separate subcomponentseach of which has its own distinctive properties that interact to yield the prop-erties of the whole. These modules are, in many cases, specific to language.Research has undermined the notion that the mind possesses only “generalprinciples of intelligence” that cover all kinds of mental activity. One moduleof innate linguistic capacity contains abstract structures that are compositional(consisting of units made up of smaller units) and that fit a narrow range ofpossibilities. Another module encompasses the ability to relate one position toanother within these structures by movement, and those movement relation-ships are narrowly defined. Another module is the mental lexicon, a list of wordforms and their crucial properties.

To see the kind of compositionality involved, consider how words combine.Words are members of categories like noun (N), verb (V), preposition (P),adjective/adverb (A). If two words combine, the grammatical properties ofthe resulting phrase are determined by one of the two words, which we call thehead; the head projects the phrase. So, if we combine the verb visit with thenoun Chicago, the resulting phrase visit Chicago has verbal and not nominalproperties. It occurs where verbs occur and not where nouns occur: I want tovisit Chicago, but not *the visit Chicago nor *we discussed visit Chicago. Sothe expression visit Chicago is a verb phrase (VP), where the V visit is the headprojecting to VP. This can be represented as a labeled bracketing (2a) or as a treediagram (2b). The verb is the head of the VP and the noun is the complement.(This is the novel, bottom-up approach to phrase structure of Chomsky [1995c:244ff.].)



(2) a. VP[Vvisit NChicago]b. VP

V Nvisit Chicago

In general, two categories merge to form a new category. So an “inflectional”element like will might merge with the VP visit Chicago, to yield the morecomplex expression will visit Chicago, with the structure of (3). The inflectionalwill heads the new phrase and projects to a phrasal category IP. This meansthat visit Chicago is a unit (VP), which acts as the complement of will, but willvisit is not a unit; that is, there is no single node which dominates will visit andnothing else in this example.

(3) IP[Iwill VP[visit Chicago]]

The units defined by these trees are the items that the computational oper-ations manipulate; they are the items that move and delete and that receiveindices. Non-units are not available to these operations.

Let us return now to the problem of the reduced is. A computational operationattaches the reduced is to the preceding word as a “clitic.” This makes ’s asintegral a part of that word as a plural-marking -s. Just as the plural -s ispronounced differently in cats, dogs, and palaces, similarly the reduced isin Pat’s here, Doug’s sad and Alice’s happy. However, a silent, understoodelement also attaches clitic-like to a host to the left and the host must be afull phonological word. In an expression like Kim is happier than Tim is, thereis no silent, understood element following the first is, which therefore maybe reduced. The fact that there is a silent, understood element following thesecond is (“happy”) means that it must be a full, phonological word and maynot be reduced.1 The same holds for ’ve, ’re, and the reduced forms of am, will,would, shall, and should. Poets make linguistic jokes from these principles: theGershwins were famous for contraction jokes and in Girl Crazy (1930) a chorusbegins I’m bidin’ my time /’Cause that’s the kind of guy I’m.

Is may be reduced in (4a), where there is no silent, understood element thatneeds a host, but not in I wonder what the problem is in Washington, whichhas the structure of (4b). Here what has moved from the position indicated, isunderstood in the position x, and reduction is not possible. So now we have ananswer to the problem sketched at the outset: a reduced is is a clitic and maynot host a silent, understood element, which also has clitic-like properties.

(4) a. The problem is in Washington.b. I wonder whatx the problem is x in Washington.


48 David Lightfoot

Clitics are little words which occur in many, perhaps all languages, and havethe property of not being able to stand alone. Part of what a child growing agrammar needs to do is to determine the clitics in his or her linguistic environ-ment, knowing in advance of any experience that these are small, unstresseditems attached to an adjacent element. This predetermined knowledge (which Idescribe here informally) is contributed by the linguistic genotype and is whatthe child brings to language acquisition. So hearing a reduced form like It’scold in here and knowing that it is equivalent in some way to It is cold in heresuffices to show the child that ’s is a clitic; the child also knows in advance ofany experience that understood elements require an appropriate phonologicalhost, a full phonological word.

Under this approach, the child is faced with a chaotic environment and scansit, looking for clitics . . . among many other things, of course (Lightfoot 1999).This is the answer that we provide to our initial problem and it is an answer ofthe right shape. It makes general claims at the genetic level (clitics and theirbehavior are predefined; understood items are hosted by full, phonologicalwords) and postulates that the child arrives at a plausible analysis on exposureto a few simple expressions like It’s cold in here. The analysis that the childarrives at predicts no reduction for the underlined is in Kim is happier than Timis, I wonder what the problem is in Washington, and countless other cases, andthe child needs no correction in arriving at this system. The very fact that ’s isa clitic, defined in advance of any experience, dictates that it may not occur incertain contexts, where it would host an understood item. It is for this reasonthat the generalization that is may be pronounced as ’s breaks down at certainpoints and does not hold across the board, and the analysis must generalize toclitics in all languages.

Consider now the second problem, the reference of pronouns. They taughtus in school that pronouns refer to a preceding noun, but the data of (1) showthat that isn’t always right. As we saw, in (1d) him may not refer to Jay; in (1b)him may refer to Jay but not to Jay’s brother. The best account of this complexphenomenon seems to be to invoke a native principle which says that pronounsmay not refer back to a local nominal element, where “local” means containedin the same clause (IP) or in the same noun phrase (NP).

In (5) I give the relevant structure for the corresponding sentences of (1).In (5b) the NP Jay’s brother is local to him and so him may not refer back tothat NP – we express this by indexing them differently. On the other hand, Jayis contained inside the NP and therefore is not available to him for indexingpurposes, so those two nouns do not need to be indexed differently – theymay refer to the same person and they may be coindexed.2 Again we see theconstituent structure illustrated earlier playing a central role in the way in whichthe indexing computations are carried out. In (5d) Jay is local to him and sothe two elements may not be coindexed; they do not refer to the same person.



In (5c) Jay is not local to he, because the two items are not contained in thesame clause: Jay and he may refer to the same person or to different people. In(5a) his is contained inside an NP and may not be coindexed with anything elsewithin that NP; what happens outside the NP is not systematic; so his and Jaymay corefer and do not need to be indexed differently.

(5) a. IP[Jayi hurt NP[hisi/j nose]]b. IP[NP[Jayi’s brother]k hurt himi/j]c. IP[Jayi said IP[hei/j hurt Ray]]d. IP[Jayi hurt himj]

We could have illustrated this principle equally well with data from Frenchor from Dutch, because the principle applies quite generally to pronouns in alllanguages. If we assume a native principle, available to the child independentlyof any actual experience, language acquisition is greatly simplified. Now thechild does not need to “learn” why the pronoun may refer to Jay in (5a) or (5b,c)but not in (5d). Rather, the child raised in an English-speaking setting has onlyto learn that he, his, him are pronouns, i.e. elements subject to our principle.This can be learned by exposure to a simple sentence like (1d/5d), uttered in acontext where him refers to somebody other than Jay.

One way of thinking of the contribution of the linguistic genotype is to view itas providing invariant principles and option-points or “parameters” (Chomsky1981a,b). There are invariant principles that understood elements cliticize on to afull phonological word and that pronouns are not locally coindexed. Meanwhile,there are options that direct objects may precede the verb in some grammars(German, Japanese) and may follow it in others (English, French), that someclitics attach to the right and some to the left. These are parameters of variationand the child sets these parameters one way or another on exposure to herparticular linguistic experience. As a result, a grammar emerges in the child,part of the linguistic phenotype. The child has learned that ’s is a clitic andthat her is a pronoun; the genotype ensures that an understood item is notattached to ’s and that her is never used in a context where it refers to a localnominal.

Here we have looked at two specific acquisition problems and consideredwhat ingredients are needed for their solution. Now let us stand back and thinkabout these matters more abstractly.

The acquisition problem: the poverty of the stimulus

The child acquires a finite system, a grammar, which generates structures thatcorrespond more or less to utterances of various kinds. Some structural principleprevents forms like *Kim’s happier than Tim’s from occurring in the speech ofEnglish speakers, as we have seen. Children are not exposed to pseudosentences


50 David Lightfoot

like this and informed systematically that it is not said. Speakers come to knowsubconsciously that it cannot be said and this knowledge emerges somehow,even though it is not part of the input to the child’s development. It is notenough to say that people do not utter such forms because they never hearthem. This argument is insufficient because people say many things that theyhave not heard, as we have noted. Language is not learned simply by imitatingor repeating what has been heard.

This poverty-of-stimulus problem, Plato’s Problem, defines our approach tolanguage acquisition. Over the last forty years, much of the linguistic literaturehas focused on areas where the best description cannot be derived directly fromthe data to which the child has access, or is underdetermined by those data. Ifthe child’s linguistic experience does not provide the basis for establishing aparticular aspect of linguistic knowledge, there must be another source for thatknowledge.

This is not to say that imitation plays no role, just that it does not provide asufficient explanation. Nobody denies that the child must extract informationfrom her environment; it is no revelation that there is “learning” in that technicalsense. Our point is that there is more to language acquisition than this. Childrenreact to evidence in accordance with specific principles. Learning of the kind thattakes place in schools or in psychologists’ laboratories involves generalization,association, induction, conditioning, hypothesis forming and testing, etc., butthis does not play any significant role in explaining children’s acquisition oflanguage.

The problem demanding explanation is compounded by other factors. Despitevariation in background and intelligence, people’s mature linguistic capacityemerges in fairly uniform fashion, in just a few years, without much apparenteffort, conscious thought, or difficulty; and it develops with only a narrowrange of the logically possible “errors.” Children do not test random hypotheses,gradually discarding those leading to “incorrect” results and provoking parentalcorrection. In each language community the non-adult sentences formed by veryyoung children seem to be few in number and quite uniform from one child toanother, which falls well short of random hypotheses. Normal children attaina fairly rich system of linguistic knowledge by five or six years of age and amature system by puberty.3 In this regard, language is no different from, say,vision, except that vision is taken for granted and ordinary people give moreconscious thought to language.

These, then, are the salient facts about language acquisition, or more properly,language growth. The child masters a rich system of knowledge without sig-nificant instruction and despite an impoverished stimulus; the process involvesonly a narrow range of non-adult forms and it takes place rapidly, even explo-sively between two and three years of age. The main question is how childrenacquire so much more than they experience.



A grammar represents what a speaker comes to know, subconsciously for themost part, about his or her native language. It represents the fully developedlinguistic capacity, and is therefore part of an individual’s phenotype. It is oneexpression of the potential defined by the genotype. Speakers know what aninfinite number of sentences mean and the various ways in which they canbe pronounced and rephrased. Most of this largely subconscious knowledgeis represented in a person’s grammar. The grammar may be used for variouspurposes, from everyday functions like expressing ideas, communicating, orlistening to other people, to more contrived functions like writing elegant proseor lyric poetry, or compiling and solving crossword puzzles, or writing an articleabout Plato’s problem.

I do not want to give the impression that all linguists adopt this Chom-skyan view of things. People have studied language with quite different goalsin mind, ranging from the highly specific (to describe Dutch in such a waythat it can be learned easily by speakers of Indonesian), to more general goals,such as showing how a language may differ from one historical stage to another(comparing, say, Chaucerian and present-day English). However, the researchparadigm sketched here (and in Chomsky 1959, 1965, 1975, etc.) has been thefocus of much activity over the last forty years and it construes a grammar as abiological object, the language organ.

The analytical triplet

A grammar, under this view, is a psychological entity, part of the psychologicalstate of somebody who knows a language. For any aspect of linguistic knowl-edge, three intimately related items are included in the account. First, thereis a formal and explicit characterization of what a mature speaker knows; thisis the grammar, which is part of that speaker’s phenotype. Since the grammaris represented in the mind/brain, it must be a finite system, which can relatesound and meaning for an infinite number of sentences.

Second, also specified are the relevant principles and parameters commonto the species and part of the initial state of the organism; these principles andparameters make up part of the theory of grammar or Universal Grammar, andthey belong to the genotype.

The third item is the trigger experience, which varies from person to personand is embedded in an unorganized and haphazard set of utterances, of the kindthat any child hears (the notion of a trigger is from ethologists’ work on theemergence of behavioral patterns in young animals and, as in those domains,consists of abstract structures, “cues” in the sense of Lightfoot 1999). Theuniversal theory of grammar and the variable trigger together form the basisfor attaining a grammar; grammars are attained on the basis of a certain triggerand the genotype.


52 David Lightfoot

(6) is the explanatory schema, with general biological terminology in (6a)and the corresponding linguistic terms in (6b). The triggering experience causesthe genotype to develop into a phenotype; exposure to a range of utterancesfrom, say, English allows the UG capacity to develop into a particular maturegrammar. One may think of the theory of grammar as making available a setof choices; the choices are taken in the light of the trigger experience or the“PLD,” and a grammar emerges when the relevant options are resolved.

(6) a. linguistic triggering experience (genotype → phenotype)b. Primary Linguistic Data (Universal Grammar → grammar)

Each of the items in the triplet – trigger, UG, and grammar – must meetvarious demands. The trigger or PLD consists only of the kinds of things thatchildren routinely experience and includes only simple structures. The theoryof grammar or UG is the one constant and holds universally such that anyperson’s grammar can be attained on the basis of naturally available triggerexperiences. The mature grammar defines an infinite number of expressionsas well-formed, and for each of these it specifies at least the sound and themeaning. A description always involves these three items and they are closelyrelated; changing a claim about one of the items usually involves changingclaims about the other two.

The conditions of language acquisition make it plain that the process must be largelyinner-directed . . . which means that all languages must be close to identical, largelyfixed by the initial state. The major research effort since has been guided by this tension,pursuing the natural approach: to abstract from the welter of descriptive complexity cer-tain general principles governing computation that would allow the rules of a particularlanguage to be given in very simple forms, with restricted variety. (Chomsky 2000a: 122)

The grammar is one subcomponent of the mind, which interacts with othercognitive capacities or modules. Like the grammar, each of the other modulesmay develop in time and have distinct initial and mature states. So the visualsystem recognizes triangles, circles, and squares through the structure of thecircuits that filter and recompose the retinal image (Hubel & Wiesel 1962).Certain nerve cells respond only to a straight line sloping downward from leftto right, other nerve cells to lines sloped in different directions. The range ofangles that an individual neuron can register is set by the genetic program, butexperience is needed to fix the precise orientation specificity (Sperry 1968). Inthe mid 1960s David Hubel, Torsten Wiesel, and their colleagues devised aningenious technique to identify how individual neurons in an animal’s visualsystem react to specific patterns in the visual field (including horizontal andvertical lines, moving spots, and sharp angles). They found that particular nervecells were set within a few hours of birth to react only to certain visual stimuli,and, furthermore, that if a nerve cell is not stimulated within a few hours, it



becomes inert. In several experiments on kittens, it was shown that if a kittenspent its first few days in a deprived optical environment (a tall cylinder paintedonly with vertical stripes), only the neurons stimulated by that environmentremained active; all other optical neurons became inactive because the relevantsynapses degenerated, and the kitten never learned to see horizontal lines ormoving spots in a normal way.

In this view, learning is a selective process: parameters are provided by thegenetic equipment and relevant experience fixes those parameters. A certainmature cognitive structure emerges at the expense of other possible structuresthat are lost irretrievably as the inactive synapses degenerate. The view thatthere is a narrowing down of possible connections out of an overabundanceof initially possible ones is now receiving more attention in the light of Hubeland Wiesel’s Nobel Prize-winning success. For the moment, this seems to be amore likely means to fine tune the nervous system as “learning” takes place, asopposed to the earlier view that there is an increase in the connections amongnerve cells.

Piattelli-Palmarini (1986, 1989) draws a helpful analogy with recent work onimmunology. The commonsense view was that the immune system developedantibodies as a kind of learning process, triggered by exposure to bacteria; anorganism would produce antibodies to counter the attack. However, Niels KajJerne won his Nobel Prize for showing that this commonsense view is incorrect:antibody formation is a selective process, not instructive. Organisms containimmense numbers of antibodies; the antigen selects and amplifies specific anti-bodies that already exist. “Looking back into the history of biology, it appearsthat wherever a phenomenon resembles learning, an instructive theory was firstproposed to account for the underlying mechanisms. In every case, this waslater replaced by a selective theory” (Jerne 1967; see also 1985).

So human cognitive capacity is made up of identifiable properties that aregenetically prescribed, each developing along one of various pre-establishedroutes, depending on the particular experience encountered. These genetic pre-scriptions may be highly specialized, as Hubel and Wiesel showed for the visualsystem. They assign some order to our experience. Experience elicits or triggerscertain kinds of specific responses but it does not determine the basic form ofthe response.

This kind of modularity is very different from the view that the cognitive fac-ulties are homogeneous and undifferentiated, that the faculties develop throughgeneral problem-solving techniques. In physical domains, nobody would sug-gest that the visual system and the system governing the circulation of the bloodare determined by the same genetic regulatory mechanisms.

Of course, the possibility should not be excluded that the linguistic princi-ples postulated here may eventually turn out to be special instances of principlesholding over domains other than language, but before that can be established


54 David Lightfoot

much more must be known about what kinds of principles are needed for lan-guage acquisition to take place under normal conditions. Similarly for otheraspects of cognitive development. Only then can meaningful analogies bedetected. Meanwhile,

we are led to expect that each region of the central nervous system has its own specialproblems that require different solutions. In vision we are concerned with contours anddirections and depth. With the auditory system, on the other hand, we can anticipate agalaxy of problems relating to temporal interactions of sounds of different frequencies,and it is difficult to imagine that the same neural apparatus deals with all of thesephenomena . . . for the major aspects of the brain’s operation no master solution islikely. (Hubel 1978: 28)

Real-time acquisition of grammars

In the domain of language, ingenious colleagues at the University of Marylandhave shown that the distinctions discussed at the beginning of this chapter donot result from learning and that the hypothesized genetic constraints seem tobe at work from the outset. The experimenters constructed situations in whichchildren would be tempted to violate the relevant constraints. The fact thatchildren conform to the hypothesized constraints, resisting the preferences theyshow in other contexts, is taken to be evidence that they have the constraintsunder investigation and they have them at the earliest stage that they might bemanifested (Crain 1991).

Stephen Crain and Rosalind Thornton developed an elicitation task thatencouraged children to ask questions like *Do you know what that’s up there?They hypothesized that children would generally show a preference for thereduced ’s form whenever this was consistent with their grammars. This prefer-ence is revealed in a frequency count of legitimate forms, like Do you know whatthat’s doing up there? Comparing the frequency of the reduced forms in thesecontexts with non-adult reduced forms would indicate whether or not child-ren’s grammars contained the hypothetical genetic constraint. If the geneticconstraint is at work, there should be a significant difference in frequency;otherwise, not.

Thornton and Crain conducted an experiment to elicit a long-distance ques-tion. The target productions were evoked by the following protocols.

Protocols for rightward cliticization

(7) Experimenter: Ask Ratty if he knows what that is doing up there.Child: Do you know what that’s doing up there?Rat: It seems to be sleeping.



(8) Experimenter: Ask Ratty if he knows what that is up there.Child: Do you know what that is up there?Rat: A monkey.

In (7) the child is invited to produce a sentence where what is understood asthe object of doing: do you know whatx that is doing x up there? Therefore, ismay be cliticized because it is not needed to host the silent, understood item x;x is hosted by doing. However, in (8) the child produces a sentence where whatis understood as the complement of is, i.e. between is and the following item:do you know whatx that is x up there? (cf. That is a bottle up there). Here isis needed in its full form in order to host x; no adult would say *Do you knowwhat that’s up there?, with the reduced form (cf. That’s a bottle up there).

Thornton and Crain found that young children behaved just like adults,manifesting the hypothetical genetic constraint. The children tested ranged inage from 2 years, 11 months to 4 years, 5 months, with an average age of3 years, 8 months. In the elicited questions there was not a single instance ofthe reduced form where it is impossible in adult speech. Children producedelaborate forms like those of (9), but never with the reduced form of is.

(9) a. Do you know what that black thing on the flower is? (4 years,3 months)

b. Squeaky, what do think that is? (3 years, 11 months)c. Do you know what that is on the flower? (4 years, 5 months)d. Do you know what that is, Squeaky? (3 years, 2 months)

There is, of course, much more to be said about grammars and their acqui-sition, and there is an enormous technical literature (Crain & Thornton 1998).Meanwhile, we have an approach to the riot of differences that we find in thelanguages of the world and even within languages. As children, our linguisticexperience varies tremendously; no two children experience the same set ofsentences, let alone the same pronunciations. Nonetheless, the approach wehave sketched enables us to understand the universality of our development,why we categorize the linguistic world so similarly and can talk to each otherdespite variation in childhood experience.

The organ

So the human capacity for natural language results from – and is made possibleby – a biologically determined organ specific both to this domain and to ourspecies. Efforts to teach human languages to individuals of other species, eventhose closest to us, have uniformly failed. While a certain capacity for arbitrarysymbolic reference can be elicited in some higher apes (Premack 1980, 1990;Premack & Woodruff 1978) and perhaps even in other animals, human-type


56 David Lightfoot

syntactic systems are well beyond the capacity of non-humans . . . just ashumans, even with intensive training, are incapable of free flight.

The functional properties of our language capacity develop along a regu-lar maturational path, such that it seems more appropriate to see our linguisticknowledge as “growing” rather than being “learned.” As with the visual system,much of the detailed structure we find is “wired in,” though triggering experi-ence is necessary to set the system in operation and to determine some of itsspecific properties. The deep similarity among the world’s languages supportsthe notion that they are the product of a common human faculty. The manuallanguages which develop in Deaf communities independently of one anotheror of the language of the surrounding hearing community share in these fun-damental properties. The profound structural similarities between signed andspoken languages, including not only the basic principles of organization butthe specific path of their development, the brain regions associated with theircontrol, and other factors, are neither the result of shared history nor sharedproperties of gestural and articulatory/acoustic/auditory modalities, but ratherthey derive from shared biology (Newport 1999; Supalla 1990).

The development of structurally deficient pidgins into the essentially normallinguistic systems found in creoles supports the richness of the genotypic sys-tem involved in linguistic development. Pidgins change significantly as theybecome creoles and this is an automatic result of transmission through the nat-ural language acquisition process in new generations of children (Bickerton1999; DeGraff 1999; Lefebvre 1998). Less dramatic linguistic changes fromone generation to another within the history of, say, English can also be under-stood sometimes through ideas of the linguistic genotype. Certain aspects oflanguage change take place in fits and starts and that characteristic bumpinessof change can be understood as a matter of parameters being reset at certaincritical points in linguistic history (Lightfoot 1999).

The language faculty has properties typical of a bodily organ, a specializedstructure destined to carry out a particular function. Some organs, like the bloodand the skin, interact with the rest of the body across a widespread, complexinterface, and all organs are integrated into a complex whole. Often the limitsto an organ are unclear, and anatomists do not worry about whether the hand isan organ or one of its fingers. It is clear that the body is not made up of creamcheese, and the same seems to be true of the brain.

The language organ is not to be interpreted as having an anatomical localiza-tion comparable to that of the kidney, at least not at this stage of knowledge. Ourunderstanding of the localization of cognitive function in brain tissue is muchtoo fragmentary and rudimentary to support a claim along those lines. Certaincortical and subcortical areas can be shown to subserve functions essential tolanguage, in the sense that lesions in these areas disrupt language functioning(sometimes in remarkably specific ways), but an inference from this evidence



to a claim that “language is located in Broca’s (and/or Wernicke’s) area” isunwarranted. Even the claim that language functions are located in the leftcortical hemisphere seems to be an oversimplification (Kosslyn et al. 1999).At this stage, linguistic capacity is better understood in functional rather thananatomical terms, along the lines that I have indicated. Even if it were to emergethat there is no clear distinction between language-related and non-language-related brain tissue, it would still be useful to treat the language capacity as adiscrete human biological system in functional terms.

The domain-specificity of the language faculty is supported by the manydissociations that have been observed between control of language structureand other cognitive functions. If a system operates independently of other sys-tems, it is a candidate for modular status. So with the senses, one can be deafwithout being blind, and vice versa, which supports the claim that hearing andsight are products of distinct systems. Neil Smith (2003) provides an excellentdiscussion of this point. He discusses a linguistic savant, Christopher. Christo-pher’s hand–eye coordination is severely impaired and his psychological profileshows “moderate to severe disability in performance tasks, but results close tonormal in verbal tasks.” Despite low general intelligence, he has an astonishingcapacity to pick up languages; see also Smith & Tsimpli (1995). Some kindsof aphasia show the reverse, and likewise Specific Language Impairment (SLI;for an overview, see Joanisse & Seidenberg 1998). SLI children are cognitivelynormal but fail to develop age-appropriate linguistic capacities (Bishop 1997).Researchers have postulated a range of grammatical deficits (Clahsen et al.1997; Gopnik 1997; van der Lely 1996) and Levy & Kave (1999) offer a usefuloverview.4

Smith points to other dissociations:

Just as intelligence and language are dissociable, so also is it possible to separate linguis-tic ability and Theory of Mind, with autistic subjects lacking in the latter but (potentially,especially in the case of Asperger’s Syndrome [Frith 1991]) language being retainedwithin normal limits. Some Down Syndrome children provide a contrary scenario, withtheir Theory of Mind being intact, but their linguistic ability moderately to severelydegraded. (2003)

Similarly we find “submodular” dissociations within the language organ,suggesting that grammars have their own internal modules. Smith points todissociations between the lexicon and the computational system. Christopher’stalent for learning second languages “is restricted largely to mastery of themorphology and the lexicon, whilst his syntactic ability rapidly peaks and thenstagnates . . . [A] reverse dissociation [is] found in the case of children withSpinal Muscular Atrophy, who seem to develop a proficient syntactic rule sys-tem but have correspondingly greater difficulty with lexical development (seeSieratzki & Woll 2002)” (Smith 2003). Edwards & Bastiaanse (1998) address


58 David Lightfoot

this issue for some aphasic speakers, seeking to distinguish deficits in the com-putational system from deficits in the mental lexicon.

We also know that focal brain lesions can result in quite specific languageimpairments in the presence of normal cognitive abilities, and vice versa.Friedmann & Grodzinsky (1997) argue that agrammatic aphasics may lackcertain functional categories. Ingham (1998) describes a young child in similarterms, arguing that she lacked one particular functional category.

This modular view runs contrary to a long tradition, often associated withJean Piaget, which claims that language is dependent on prior cognitive capac-ities and is not autonomous and modular (Piaget & Inhelder 1968; Piattelli-Palmarini 1980 for critical discussion). This claim is undermined by the kindsof dissociations that have been observed. Bellugi et al. (1993) have shown, foranother example, that Williams Syndrome children consistently fail to pass seri-ation and conservation tests but nonetheless use syntactic constructions whoseacquisition is supposedly dependent on those cognitive capacities.

Conclusion

Recent theoretical developments have brought an explosive growth in whatwe know about human languages. Linguists can now formulate interestinghypotheses, account for broad ranges of facts in many languages with elegantabstract principles. We understand certain aspects of language acquisition inyoung children and can model some aspects of speech comprehension.

Work on human grammars has paralleled work on the visual system and hasreached similar conclusions, particularly with regard to the existence of highlyspecific computational mechanisms. In fact, language and vision are the areas ofcognition that we know most about. Much remains to be done, but we can showhow children attain certain elements of their language organs by exposure toonly an unorganized and haphazard set of simple utterances; for these elementswe have a theory which meets basic requirements and we have, at some level,a solution to aspects of Plato’s Problem. Eventually, the growth of language in achild will be viewed as similar to the growth of hair: just as hair emerges with acertain level of light, air, and protein, so, too, a biologically regulated languageorgan necessarily emerges under exposure to a random speech community.

From the perspective sketched here, our focus is on internal, individualgrammars (“I-language” in the terminology of Chomsky 1986), not on theproperties of a particular language or even of general properties of many orall languages. A language under this view is an epiphenomenon, a derivativeconcept, the output of certain people’s grammars (perhaps modified by othermental processes). A grammar is of clearer status: the finite system that charac-terizes an individual’s linguistic capacity and is represented in the individual’smind/brain. No doubt the grammars of two individuals whom we regard as



speakers of the same language will have much in common, but there is noreason to worry about defining “much in common,” about specifying when theoutputs of two grammars constitute one language. Just as it is unimportant formost work in molecular biology whether two creatures are members of the samespecies (as emphasized, for example by Monod [1972: ch. 2] and by Dawkins[1976]), so too the notion of a language is not likely to have much importancewithin this biological perspective. Of course, there is more to the study of lan-guage than the biological perspective provides and other lenses focus differentlyon different phenomena.


3 Grammar, levels, and biology

Howard Lasnik

Introduction

Language is an immensely rich phenomenon, presenting vast challenges forthe linguist, the scientist of this phenomenon. Among the most central, andmost difficult, questions are ones concerning the nature of the capacity we allhave to speak a language. Just what is this capacity, and how does it arise inthe individual? We thus have (among many others) the two following relatedquestions:– What is the correct characterization of someone who “knows a language”

(in general terms, who has command of a systematic connection betweensound and meaning)?

– How does that systematic connection arise in the individual?For the first of these questions, the linguist hopes to account, in an explicit

way, for the speaker’s ability to put together and understand sentences, includingones new to the speaker (and often new to all speakers), and for the speaker’sability to judge potential sentences as “acceptable” (John left) or “unacceptable”(Left John). For the second question, a particularly difficult one given howcomplex the capacity seems to be and how quickly it is acquired, the linguistseeks to discover what aspects of the capacity are determined by the child’sexperience, and how this determination (“language learning”) takes place. Fora half century, Noam Chomsky has been developing a theory of language thatdeals with these two questions, by positing explicit formulations of humanlanguage capacity in terms of a productive “computational” system, most ofwhose properties are present in advance of experience, “wired in” in the structureof the human brain. Thus, Chomsky conceives of his enterprise as part ofpsychology, ultimately biology.

From Noam Chomsky’s earliest theories of language to his most recent, sim-plicity has been a guiding concern, pervading all of his analyses. Throughout,this drive towards simplicity has had two major motivations. First, there is thestandard, virtually universal, gamble that scientists make: that the world, andthe portion of it they are investigating, is simple. Second, in the case of a cog-nitive system like language, the assumption of simplicity at least helps provide

60


Grammar, levels, and biology 61

the basis for answering the question of how the part of the system that must beacquired is acquired.

Chomsky (1965) proposes criteria of adequacy for theories of language thatrelate to how successfully these two questions are answered. “Descriptiveadequacy” is the criterion for hypothesized answers to the first question(“grammars”), as explicated in the following passage:

A grammar can be regarded as a theory of a language; it is descriptively adequate to theextent that it correctly describes the intrinsic competence of the idealized native speaker.The structural descriptions assigned to sentences by the grammar, the distinctions thatit makes between well-formed and deviant, and so on, must, for descriptive adequacy,correspond to the linguistic intuition of the native speaker (whether or not he may beimmediately aware of this) in a substantial and significant class of crucial cases. (1965:24)

The second question concerns “explanatory adequacy”:

To the extent that a linguistic theory succeeds in selecting a descriptively adequategrammar on the basis of primary linguistic data [the information available to the childin the process of language acquisition HL], we can say that it meets the condition ofexplanatory adequacy. (1965: 25)

The learner’s task is seen as choosing the correct grammar from all the (biolog-ically) possible ones. This means that the linguist’s model of the component ofthe human mind concerned with language acquisition (the “language acquisi-tion device”) must show how the correct grammar is selected from among thepossible ones.

Generative grammar as a theory of language

Chomsky called his approach “generative grammar.” By the term “generative,”borrowed from mathematics, he meant simply that he intended to formulateexplicit answers to the questions outlined above.1 The particular explicit answerChomsky provided was in terms of the mathematical formalism of set theory.Any remotely adequate theory will have to account for the fact that there isno limit on the number of sentences in any human language. Below, we willconsider two attempts to deal with this property.

Already in his first detailed piece of linguistic analysis, Chomsky (1951)was explicitly concerned with descriptive adequacy, even though not under thatterm:

A grammar of a language must . . . correctly describe the “structure” of the language (i.e.,it must isolate the linguistic units, and, in particular, must distinguish and characterize justthose utterances which are considered “grammatical” or “possible” by the informant . . .(1951: 1)


62 Howard Lasnik

And explanatory adequacy (though again, not under that term) is fundamentalin Chomsky’s magnum opus, The Logical Structure of Linguistic Theory (LSLT,1955):

We are antecedently interested in developing a theory that will shed some light on suchfacts as the following:

1 A speaker of a language has observed a certain limited set of utterances in his language.On the basis of this finite linguistic experience he can produce an indefinite number ofnew utterances which are immediately acceptable to the speech community. He can alsodistinguish a certain set of “grammatical” utterances, among utterances that he has neverheard and might never produce. He thus projects his past linguistic experience to includecertain new [sentences] while excluding others. (1955: 61)

That is, based on finite, and quite limited, input, the child creates a productivesystem that goes far, indefinitely far, beyond that data.

Explanatory adequacy in Chomsky’s sense concerns language acquisition.Theories that seek to attain explanatory adequacy must posit some innate struc-ture in the mind. This is surely indisputable; while a human being can learnlanguage, a rock, or a gerbil, cannot. The research question, ultimately a ques-tion of biology, concerns just what this innate structure is. Chomsky (1965)poses the issue this way:

To learn a language . . . the child must have a method for devising an appropriate grammar,given primary linguistic data. As a precondition for language learning, he must possess,first, a linguistic theory that specifies the form of the grammar of a possible humanlanguage, and, second, a strategy for selecting a grammar of the appropriate form that iscompatible with the primary linguistic data. As a long-range task for general linguistics,we might set the problem of developing an account of this innate linguistic theory thatprovides the basis for language learning. (1965: 25)

This task became more and more central in Chomskyan work of the 1970s,1980s, and 1990s.

On first examination, human languages appear to be almost overwhelminglycomplex systems, and the problems, for the linguist, of successfully analyzingthem, and for the learner, of correctly acquiring them, virtually intractable. Butif the system is broken down into smaller parts, the problem might likewise bedecomposed into manageable components. In fact, on this divide and conquer(“modular”) approach, by the 1980s, languages began to seem much simpler,as I will discuss later. To illustrate modularity with a few simple examples, itis uncontroversial that an utterance is made up of sounds (or gestures moregenerally when we include signed languages). However, while the acousticsignal itself is continuous, it turns out to be crucial to analyze it in termsof discrete elements, “phones,” in order to capture basic regularities. Further,analysis into phones does not suffice for linguistic description. An aim anda name are phonetically indistinguishable, but are quite different linguisticevents. Analysis of the sequence of phones into words is thus necessary. But



even analysis into a sequence of words does not, in general, capture all thesalient properties of an utterance in a human language. Consider the followingfamous example, first discussed by Chomsky in the 1950s:

(1) Flying planes can be dangerous.

A moment’s reflection reveals that this one string of words is two differentsentences, a phenomenon Chomsky called “constructional homonymity” andthat is now generally called “structural ambiguity.” The two sentences combinedin (1) have the following paraphrases:

(2) They (the objects) can be dangerous.

(3) It (the activity) can be dangerous.

Structure, the categorization and grouping of the words, is necessary if we areto provide an adequate description of such an example as (1). A similar exampleis given in (4):

(4) Mary saw the man with binoculars.

The seeing is with binoculars, or the man has binoculars. The required charac-terizations of a sentence in all these different terms (including phones, words,phrase structure) are at the heart of Chomsky’s theory of levels:

A language is an enormously complex system. Linguistic theory attempts to reducethis immense complexity to manageable proportions by the construction of a systemof linguistic levels, each of which makes a certain descriptive apparatus available forthe characterization of linguistic structure. A grammar reconstructs the total complex-ity of a language stepwise, separating out the contributions of each linguistic level.(1955: 63)

Chomsky sometimes referred to these levels as “levels of representation,” aname that became standard, so I will use it here. But it should be pointed outthat the name can be misleading, since its use inevitably leads to confusion withthe standard philosophical notion “representation” which concerns a relationbetween linguistic expressions and portions of the world. Chomsky’s levels ofrepresentation are completely language internal. Perhaps a more accurate termmight have been “levels of structure.”

Levels of representation in a theory of language

Levels of representation fit into the theory in the following way:

We define, in general linguistic theory, a system of levels of representation. A levelof representation consists of elementary units (primes), an operation of concatena-tion by which strings of primes can be constructed, and various relations defined onprimes, strings of primes, and sets and sequences of these strings. Among the abstractobjects constructed on the level L are L-markers that are associated with sentences. The


64 Howard Lasnik

L-marker of a sentence S is the representation of S on the level L. A grammar for a lan-guage, then, will characterize the set of L-markers for each level L and will determinethe assignment of L-markers to sentences. (1955: 5–6)

Phones are the primes at one level, morphemes at another, words at yet another,and so on.

The child learning a language is assumed to bring knowledge of the levelsto bear on the task of learning. That is, the child must learn properties of thelanguage at each level, but knows the levels in advance, hence, knows what tolook for. The levels are part of “Universal Grammar,” a wired-in part of thelanguage acquisition device that constitutes part of a human being’s geneticendowment. Of course, the linguist does not know in advance of research whatthe levels are. Determining them is a scientific question, one of biologicalpsychology. Chomsky has devoted considerable attention to determining justwhat the levels of representation are in the human language faculty. In LSLT, thelevels were considered to be phonetics, phonemics, word, syntactic category,morphemics, morphophonemics, phrase structure, transformations. Each wasmotivated by at least an informal argument. The “interface” level, phonetics,was justified by an argument from simplicity, a recurrent theme in Chomsky’swork from the 1950s to the present: the characterization of the levels is simplifiedif we posit a “lowest” level of phonetics Pn, whose primes are phonetic symbols,which ultimately relate to the acoustic and articulatory properties of the soundsof human language.

Although LSLT was concerned most centrally with syntax, Chomskyformulates a complete theory of grammar from sound “upwards.” And, in fact,his contributions to phonology (the component of language closest to sound)were arguably just as influential in the 1950s and 1960s as his contributions tosyntax. In LSLT, he formulated a model that incorporates phonology in a spe-cific way. In particular, he proposed that the next “higher” level above phoneticsis phonemic representation, described as follows:

The first really significant linguistic level is the level Pm of phonemes. The essence ofphonemic theory is contained in the definition of the mapping φpm that carries strings ofphonemes into strings of phones. (1955: 159)

There are classic arguments for this level, which is related to sound but is moreabstract than the level of phonetics. One class of arguments involves situationswhere two clearly distinct phones must be treated as if they were the same formany purposes. Consider, for example, the regular English noun plural ending.It is sometimes realized as [z], as in the word sons. But it is also realized as[s], as in books. Nonetheless, the ending is felt as the same item. Further, therewould be a tremendous loss of generality if the two forms were kept completelydistinct, since their occurrence is completely predictable. When the noun endswith a sound with the phonetic property “voiced,” the plural ending is realized



as [z] (itself a voiced sound). When the noun ends with a voiceless sound, theplural ending is [s], the voiceless variant of [z]. Thus, the generalization cannotbe stated in Pm or in Pn, but rather in the mapping from Pm to Pn. In Pm,the plural ending is uniform, even though Pn makes a distinction. This typeof mapping, “assimilation” (voicing assimilation in this instance), is extremelycommon in the languages of the world.

The converse situation also arises, where one phone clearly has two distinctsources. A classic example, frequently discussed by Chomsky, concerns thepair of words writer and rider (see also Dresher, this volume). In most Englishdialects, the central consonant in both is the same, an alveolar flap usuallysymbolized as [D]. However, the former word obviously relates to write, with afinal [t], while the latter relates to ride, with a final [d]. Phonemically, we have/t/ and /d/; the distinction is neutralized in the mapping from Pm to Pn.

One very important debate in the early days of generative grammar was overa question of levels: is there a level of representation intermediate betweenphonetics and phonemics? Chomsky, following and extending the argumentof Halle (1959), concluded that there is not. Chomsky (1964) observed thatmuch work in structuralist phonology maintained, either implicitly or explic-itly, a level of representation that he labeled “taxonomic phonemics.” This levelis required to meet four conditions, one of which Chomsky calls “biunique-ness.” Biuniqueness demands a one-to-one correspondence between phonemesand phones, thus, in effect, demanding a high degree of what might be called“concreteness in phonological representations.” The substance of Halle’s argu-ment is that one seemingly simple and general phonological process of Russianwould have to be split into two separate phonological rules if the intermedi-ate level demanding biuniqueness is postulated. As Chomsky observes (1964:90), “the only effect of assuming that there is a taxonomic phonemic level isto make it impossible to state the generalization.” Chomsky points out simi-lar facts from other languages as well, concluding (1964: 91) that “the effectof the biuniqueness condition is to complicate the grammar, that is to pre-vent it from achieving descriptive adequacy.” As noted above, an appeal tosimplicity was and is a significant factor in Chomsky’s arguments for partic-ular theories and analyses. And if a certain degree of abstractness is the costof simplicity, Chomsky has always been immediately willing to accept thatcontract.

Arguments from simplicity can cut both ways. The preceding argument con-cluded that a particular level should not be posited. But, as noted above, inLSLT Chomsky also presented simplicity arguments for particular levels. Onesuch case, not yet discussed, involves morphophonemic representation. In thecase of taxonomic phonemics, the argument was that we would need to positextra rules if we assumed such a level. For morphophonemics, the argument isvirtually the reverse:


66 Howard Lasnik

We may introduce a level of representation solely because it enables us to replace a greatmany rules by a single rule about a single element of this new level. Morphophonemicrepresentation provides a good example of this. Suppose that English sentences arerepresented in terms of morphemes and phonemes. Instead of associating with eachmorpheme a set of phoneme strings, . . . along with the conditions dictating the occurrenceof each, it is often possible to rewrite each morpheme as a string of invented elementscalled morphophonemes in such a way that a relatively small number of statements aboutthe phonemic forms that these assume in various contexts will suffice to determine manyconversions of morphemes into phonemic representation. (1955: 114–15)

Chomsky’s illustration of this point centers on a semi-regular pluralizationprocess in English, by which a stem-final [f] in the singular alternates with [v]in the plural (as in wife–wives). Thus, we might describe these facts with rulessuch as the following (where � is the symbol for concatenation and pl is theplural morpheme):

(5) a. wife � pl → /wayv/ � pl (ultimately, “wives”)b. wife � X → /wayf/ � X (where X �= pl)

(6) a. leaf � pl → /liyv/ � pl (ultimately, “leaves”)b. leaf � X → /liyf/ � X (where X �= pl)etc.

Chomsky argues that a much simpler description can be obtained in terms ofmorphophonemes, in particular, the morphophoneme F. The morphophonemicrepresentations of wife and leaf are as in (7) and (8) respectively.2

(7) w � a � y � F

(8) l � i � y � F

The morphophonemes can be converted into phonemes by such rules as:

(9) a. F � pl → /v/ � pl

(10) b. F � X → /f/ � pl

The simplification is that the relationship is stated once and for all, rather thanfor each individual word that partakes in the alternation.3

Levels of representation in syntax

Chomsky maintains that this line of reasoning extends to syntax:

The same kind of argument can be used to motivate and justify the introduction of a levelof phrase structure. If there were no intervening representations between Sentence andwords, the grammar would have to contain a vast (in fact, infinite) number of conversionsof the form Sentence → X, where X is a permissible string of words. However, we findthat it is possible to classify strings of words into phrases in such a way that sentencestructure can be stated in terms of phrase strings, and phrase structure in terms of wordstrings, in a rather simple way. (1955: 115–16)



The level P of phrase structure differs from the “lower” (sound-related) levelsconsidered thus far in significant respects. First, markers at those lower levels aresimply strings of symbols. P-markers, on the other hand, must be sets of strings;the appropriate set of strings, as will be seen immediately below, provides thenecessary structural information about a sentence. The phrase structure of agiven sentence clearly cannot be adequately represented by just a string ofphones or phonemes. But just as clearly, a string of words or morphemes issimilarly insufficient, as already discussed above in connection with (1). Nowconsider even a very simple sentence such as (11).

(11) The woman studied the book.

All the processes that depend on the structure of a sentence (semantic, phono-logical, and syntactic processes) demand to know more than that (11) is a stringof the five words The, woman, studied, the, and book, in that order; or that it isa string of the six morphemes – the, woman, study, past tense, the, and book,in that order. In addition, it is necessary to know that The woman comprises aunit of structure, a “constituent’; that the book also is a constituent, and one ofthe same type as The woman (call it NP [Noun Phrase]); that studied the bookis a constituent, and of a different type (call it VP [Verb Phrase]); that the twooccurrences of the are of the same type (call it Determiner); that woman andbook are of the same type (Noun); that studied is of a different type (Verb);that the entire sequence is a Sentence. The required set of strings thus includesthose in (12).

(12) The � woman � studied � the � bookSentenceNP � VPNP � V � NPDet � N � V � Det � Netc.

This set of strings constitutes the P-marker of sentence (11), given a particu-lar (oversimplified) phrase structure grammar, one specifying that a Sentenceconsists of an NP followed by a VP; a VP consists of a V followed by an NP;etc. In now standard notation:

(13) Sentence → NP VPVP → V NPNP → Det NDet → theN → woman, book, etc.V → studied, etc.

That is, VP is rewritten as (or consists of) V followed by NP, etc. A symbol thatcannot be rewritten is a terminal symbol. The string consisting solely of terminal


68 Howard Lasnik

symbols is the terminal string. P-markers are perspicuously represented asconstituent structure “trees” as in (14).

(14) Sentence

NP VP

Det N V NP

The woman studied Det N

the book

A P-marker must indicate what substrings of the terminal string are constituents,and for the ones that are, what the “labels” of the constituents are. In LSLT, thefundamental predicate characterizing these relations is the is a relation. In the(simplified) example above, the following is a relations hold:

(15) The is a Detwoman is an Nstudied is a Vthe is a Detbook is an NThe woman is an NPthe book is an NPstudied the book is a VPThe woman studied the book is a Sentence

One of the most fundamental properties of human language is its infinitude:there is no upper bound on the length of sentences, hence no upper bound onthe number of sentences. In all human languages, ever longer sentences can beconstructed by embedding one sentence inside another, as illustrated in (16).4

(16) a. Mary reads books.b. John thinks Mary reads books.c. Susan said John thinks Mary reads books.etc.

Chomsky has proposed two major distinct ways of instantiating this embeddingproperty. The first, formalized in LSLT, merged one complete phrase structureof the sort in (14) into another. This kind of operation, combining two (or more)P-markers into a new more complex “derived” P-marker, is called a “generalizedtransformation.” Metaphorically, a generalized transformation grafts onetree onto another. As I will discuss below, Chomsky has returned to this



mechanism in his recent “minimalist” work. Alongside generalized transfor-mations, there are “singulary transformations,” which apply to P-markers andderived P-markers. They apply to one tree.

Transformations as the connection between underlying andsuperficial syntactic structure

Chomsky showed how singulary transformations can explain the relatednessbetween, for example, statements and corresponding questions:

(17) a. Susan will solve the problem. ⇒ Will Susan solve the problem?b. John is visiting Rome. ⇒ Is John visiting Rome?

The members of each pair come from the same initial P-marker, with singu-lary transformations producing the divergent surface shapes. One of the greattriumphs of the analysis of such pairs in LSLT is that it was able to use thesame singulary transformation for the interrogative sentences in (17) and thesuperficially very different one in (18).

(18) Susan solved the problem. ⇒ Did Susan solve the problem?

This was a significant achievement since the relations are felt by native speak-ers to be parallel, an otherwise mysterious fact. Chomsky also showed howin numerous situations, even properties of individual sentences cannot beadequately characterized without recourse to the descriptive power of singu-lary transformations. There was, thus, considerable motivation for this newdevice relating more abstract “underlying” structures to more superficial sur-face representations. In fact, one of the major conceptual innovations in theentire theory is the proposal that a sentence has not just one structure, closelyrelated to the way it is pronounced, but an additional abstract structure (poten-tially very different from the superficial one), and intermediate structuresbetween these two. This is fundamental to all the analyses in the Chomskyansystem.

The organization of the syntactic portion of the grammar is as follows: appli-cation of the phrase structure rules creates a P-marker, or, in the case of acomplex sentence, a set of P-markers. Then successive application of transfor-mations (singularly and generalized) creates successive phrase structure rep-resentations (“derived P-markers”), culminating in a final surface representa-tion. The syntactic levels in this theory are that of phrase structure and that oftransformations, the latter giving a “history” of the transformational derivation(the successive transformational steps creating and affecting the structure). Therepresentations at these levels are the P-marker and the T-marker respectively.The final derived P-marker is the input to phonological interpretation, and theT-marker is the input to semantic interpretation.5


70 Howard Lasnik

Levels in Chomsky’s “standard theory”

Chomsky (1965), henceforth Aspects, presented a revised conception of thegrammar, based on an alternative way of constructing complex sentences, onethat Chomsky argued was an advance in terms of simplicity and explanatory ade-quacy over the one in LSLT. In the LSLT framework, the phrase structure rulesproduce simple monoclausal structures, which can then be merged together bygeneralized transformations. Generalized transformations are thus the recursivecomponent of the grammar, the one responsible for the infinitude of language.The alternative is that the phrase structure rule component itself has a recursivecharacter. Consider again the complex sentences in (16), repeated here:

(19) a. Mary reads books.b. John thinks Mary reads books.c. Susan said John thinks Mary reads books.etc.

By adding a recursive “loop” to the set of phrase structure rules in (13), wedirectly create the possibility of ever longer sentences. The rule in (20) providesthis loop.

(20) VP ⇒ V Sentence

By (13), a Sentence contains a VP; and by (20), a VP contains a Sentence. Thus,by repeated application, we can construct a sentence inside a sentence inside asentence . . . (21) is the phrase structure “tree” for (19b) given the augmentedphrase structure rules.

(21) Sentence

NP VP

N V Sentence

John thinks NP VP

N V NP

Mary reads N

books

(19c) would be similar, but with one more level of embedding, resulting fromone more application of (20).



Under this approach to sentence embedding, unlike that in LSLT, there is oneunified structure underlying a sentence prior to the operation of any transforma-tions. This structure is the result of application of the phrase structure rules and“lexical insertion transformations” which insert items from the lexicon into theskeletal structure. Chomsky argued in Aspects that this underlying structure,which he there named “deep structure,” is the locus of important generaliza-tions and constitutes a level in the sense discussed above. Chomsky’s majorarguments for this new level were that it resulted in a simpler overall theory,and at the same time it explained the absence of certain kinds of derivationsthat seemed not to occur (or at least seemed not to be needed in the descrip-tion of sentences of human languages). Taking the second of these points first,Chomsky argued that while there is extensive ordering among singulary trans-formations (situations where a derivation produces an unacceptable sentence iftwo transformations are applied in reverse order), “ there are no known cases ofordering among generalized transformation although such ordering is permit-ted by the theory of Transformation-markers” (1965: 133). Further, while thereare many cases of singulary transformations that must apply to a constituentsentence before it is embedded, or that must apply to a “matrix” sentence afteranother sentence is embedded in it, “there are no really convincing cases of sin-gulary transformations that must apply to a matrix sentence before a sentencetransform is embedded in it . . .”

As for the first argument, Chomsky claimed that the theory of transfor-mational grammar is simplified by this change, since the notions “general-ized transformation” and “Transformation-marker” are eliminated entirely. TheP-markers in the revised theory contain all of the information of those in theLSLT version, but they also indicate explicitly how the clauses are embeddedin one another, that is, information that had been provided by the embeddingtransformations and T-markers.

This change in the theory of phrase structure, which has the effect ofeliminating generalized transformations, also has consequences for the the-ory of singulary transformations. As indicated above, in the Aspects the-ory, as in LSLT, there is extensive ordering among singulary transforma-tions. In both frameworks, the set of singulary transformations was seen asa linear sequence: an ordered list. Given the Aspects modification, this listof rules applies “cyclically,” first operating on the most deeply embeddedclause, then the next most deeply embedded, and so on, working “up thetree” until they apply on the highest clause, the entire generalized P-marker.6

Thus, singulary transformations apply to constituent sentences “before” theyare embedded, and to matrix sentences “after” embedding has taken place.“The ordering possibilities that are permitted by the theory of Transformation-markers but apparently never put to use are now excluded in principle”(1965: 135).


72 Howard Lasnik

How syntax relates to semantics

An important question for any syntactic theory is how syntax relates tosemantics: what the precise connection is between form and meaning. In LSLT,the T-marker contains all of the structural information relevant to semanticinterpretation. Katz and Postal (1964) proposed a severe restriction on just howthis structural information could be accessed. In particular, they postulated thatthe only contribution of transformations to semantic interpretation is that theyinterrelate P-markers. As Chomsky puts it, (generalized) transformations com-bine semantic interpretation of already interpreted P-markers in a fixed way.In the revised theory, which came to be called the “standard theory,” the initialP-marker, now a “deep structure,” then contains just the information relevantto semantic interpretation. To summarize the model:

the syntactic component consists of a base that generates deep structures and a transfor-mational part that maps them into surface structures. The deep structure of a sentence issubmitted to the semantic component for semantic interpretation, and its surface struc-ture enters the phonological component and undergoes phonetic interpretation. The finaleffect of a grammar, then, is to relate a semantic interpretation to a phonetic representa-tion – that is, to state how a sentence is interpreted. (1965: 135–6)

To carry out this program, following Katz and Postal (1964), Chomsky pro-posed that the many seemingly “meaning-changing” optional transformationsof LSLT be replaced by obligatory transformations triggered by a marker inthe deep structure. To take one example, earlier I noted that in LSLT, simplequestions and the corresponding statements are derived from the same initialP-marker. In the revision, those initial P-markers would be very similar but notidentical. The former would contain a marker of interrogation that would bothsignal the difference in meaning and trigger the inversion that results in theauxiliary verb appearing at the front of the sentence.

At this point in the development of the theory, the model can be graphicallyrepresented as follows, with deep structure doing the semantic work formerlydone by the T-marker:

(22) Deep Structure Semantic Interpretation

Transformations(operating cyclically)

Surface Structure Phonetic Interpretation (via the “sound- related” levels of morphophonemics, phonemics, and phonetics)



There were already questions about deep structure as the sole locus of seman-tic interpretation. To take just one example, Chomsky (1957a) observed that insentences with quantifiers, the derived structure has truth conditional conse-quences. (23a) may be true while (23b) is false, for instance if one person in theroom knows only French and German, and another only Spanish and Italian.

(23) a. Everyone in the room knows at least two languages.b. At least two languages are known by everyone in the room.

In the theory of (1957a), this is not problematic, since semantic interpretationis based on the T-marker. However, in the Aspects framework, there is a prob-lem, as Chomsky acknowledges. He speculates that the interpretive differencebetween (23a) and (23b) might follow from discourse properties, rather thangrammatical ones. The problem came to loom larger and larger, leading to atheory, elaborated by Chomsky (1970), in which both deep structure and sur-face structure contribute to semantic interpretation. In this so-called “ExtendedStandard Theory” the contribution of deep structure concerns “grammaticalrelations” such as subject of and object of. The contribution of surface struc-ture concerns virtually all other aspects of meaning, including scope, as in theexamples mentioned just above, anaphora, focus, and presupposition.

Alongside these questions about deep structure as the sole locus of semanticinterpretation, there were also challenges to its very existence. Postal (1972)argued that the best theory is the simplest, which, by his reasoning, included auniform set of rules from semantic structure all the way to surface form, withno significant level (i.e. deep structure) in between.7 And McCawley (1968)explicitly formulates an argument against deep structure on the model of Halle’sargument against a level of taxonomic phonemics. McCawley’s argument isbased on the interpretation of sentences with respectively, such as (24).

(24) Those men love Mary and Alice respectively.

He argues that in a theory with deep structure, two rules, instead of one, willbe needed to determine the meaning of such sentences.

Chomsky considers this argument, but rejects it, claiming that it rests on anequivocation about exactly what the relevant rule(s) would be on the theoriesin question. Chomsky does, however, accept McCawley’s contention that it isnecessary to provide some justification for the postulation of deep structure. Buthe observes that the same is true of surface structure or phonetic representation,or, in fact, any theoretical construct. How can such a justification be provided?

There is only one way to provide some justification for a concept that is defined in termsof some general theory, namely, to show that the theory provides revealing explanationsfor an interesting range of phenomena and that the concept in question plays a role inthese explanations. (1970: 64)


74 Howard Lasnik

As far as Chomsky was concerned, this burden had been met, especially by theAspects analysis of the transformational ordering constraints discussed above.

Levels in the “Government–Binding” model

As already noted, the semantic role of deep structure was seriously diminished inthe Extended Standard Theory. It was to be diminished even further. Recall thatthe semantic contribution of deep structure in EST is limited to determination ofgrammatical relations. In deep structure, grammatical relations are transparentlyrepresented, while in derived structure, they often seem not to be represented atall. Following Chomsky (1965), let us suppose that the “subject” grammaticalrelation is assigned to an NP immediately dominated by Sentence, and theobject relation to an NP immediately dominated by VP. Consider again (19a),with the following approximate structure:

(25) Sentence

VP

N

NP

NPV

Mary reads books

The NP Mary is configurationally determined as subject and the NP booksis configurationally determined as object. But there are sentences in whichthe grammatical relations are parallel to these, but whose derived structuresseemingly do not support these determinations. “Subject-raising” sentences areone instance, and interrogatives are another:

(26) John seems to be a nice fellow. (cf. It seems that John is a nice fellow.)

(27) a. What does Mary read?b. Who does John think reads books?

In these examples, arguments appear displaced from their basic positions, thoseresponsible for their understood grammatical relations. For instance, in (27b),Who is the understood subject of reads books but in surface form it is ininitial position of the entire complex sentence. Thus, the rules determininggrammatical relations, a fundamental aspect of semantic interpretation, appar-ently need access to deep structure. However, a technical innovation in theearly 1970s obviated this dependence on deep structure interpretation. On thebasis of a variety of phenomena, Chomsky (1973a) argued that when an item



moves, it leaves behind a “trace” marking the position from which it moved.8

Given “trace theory,” Chomsky (1975) suggests that surface structure can bethe input for all semantic interpretation. Recall that the obstacle to such anapproach was that movement operations seemed to destroy the configurationsnecessary for the determination of grammatical relations. But once traces areposited, this is no longer obviously so. Chomsky summarizes the situation asfollows:

to understand the sentences we have been discussing we must surely . . . know theposition in the initial phrase marker of the phrase that has been moved. Thus consideragain [(28)], derived by NP-preposing from [(29)]:

(28) John seems [s t to be a nice fellow]

(29) seems [s John to be a nice fellow]

To understand the sentence [(28)] we must know that “John” is the subject of theembedded sentence. The initial phrase marker provides this information, but the surfacestructure (it appears) does not . . . In fact, it was precisely such considerations as thesethat motivated the principle of the standard theory that deep structures (our “initial phrasemarkers”) determine semantic interpretation.

But notice that under the trace theory of movement rules, the motivation disappears.The position of the . . . trace in surface structure allows us to determine the grammaticalrelation of “John” in [(28)] as subject of the embedded sentence. Similarly, in the othercases . . . There is a great deal [of] evidence that surface-structure information contributesto the determination of meaning. Thus, it seems reasonable to postulate that only surfacestructures undergo semantic interpretation . . . (1975: 95–6)

Suppose we use the term “Logical Form” for the syntactic representationthat interfaces with semantics. In the theory just outlined, surface structure isLogical Form (LF). However, in the late 1970s, arguments were put forward thattransformational operations of the sort successively modifying deep structure,ultimately creating surface structure, also apply to surface structure, creatingan LF that is distinct from surface structure. May (1977) argued extensively foran operation moving quantifiers from their surface positions to positions moretransparently representing their scope, with the traces of the moved quantifiersultimately interpreted as variables bound by those quantifiers. May showed howsentences with scope ambiguities receive multiple LF representations undersuch an approach. For example, a sentence like (30) has the two LF represen-tations in (31), depending on the order in which the two quantifiers are raised.Subscripts mark the association of quantifier with trace.

(30) Some student solved every problem.

(31) a. some studenti [every problemj [ti solved tj]b. every problemj [some studenti [ti solved tj]


76 Howard Lasnik

(31a) represents the reading of (30) in which some student has wider scopethan every problem, and (31b) represents the reading in which every problemhas wider scope. Note that unlike the transformational operations mentionedearlier, applications of “Quantifier Raising” (QR) exhibit no phonological dis-placement. All the expressions in (30) are pronounced in their surface-structureposition, on either reading. The model of grammar making this possible isschematized in (32), where deep structure is now “merely” the starting pointof the syntactic derivation. It has been stripped of its previous direct role insemantic interpretation.

(32) Deep Structure

Transformations

Surface Structure

PhonologicalRules

Logical Form

LF

Transformations

Phonetic FormPF

When a transformation operates between deep structure and surface structure,it will have an effect on the phonetic output, since surface structure feedsinto PF. On the other hand, a transformational application between surfacestructure and LF will have no phonetic effect, since LF does not feed intoPF. QR in (31) is an example of the latter type of “covert” transformationaloperation.

The overt movement process instantiated in (27) above, usually calledWH-movement (since interrogative expressions in English often begin withthe letters WH), was assumed to have a covert analogue as well. For instance,under the plausible assumption that overt WH-movement positions an interrog-ative operator in its natural position for interpretation (with the trace it leavesbehind in the natural position for a variable bound by the operator), in sen-tences with multiple interrogatives, such as (33), at the level of LF all are insentence-initial operator position, as illustrated in (34).

(33) Where should we put what?

(34) whati [wherej [we should put ti tj]



(34) is then rather transparently interpreted as:

(35) For which object x and which place y, we should put x at y.

Huang (1981/82, 1982) presented one of the most powerful arguments for theexistence of covert WH-movement. Chomsky observed in the early 1960s thatit is difficult to move an interrogative expression out of an embedded question(a question inside another sentence):

(36) *Whyi do you wonder [whatj [John bought tj ti]]

If (36) were acceptable, it would mean “What is the reason such that you wonderwhat John bought for that reason?” Huang showed that in Chinese, whereinterrogative expressions do not seem to move, their interpretation apparentlyobeys the same constraints that the movement in a language like English obeys.So, in Chinese an example like (37) is possible but one like (38) is impossibleon the relevant reading:

(37) ni renwei [Lisi weisheme mai-le shu]you think Lisi why bought book“What is the reason such that you think Lisi bought a book for thatreason?”

(38) (*)ni xiang-zhidao [Lisi weisheme mai-le sheme]you wonder Lisi why bought what

“What is the reason such that you wonder what Lisi bought for thatreason?”

Hence, this movement constraint seems to obtain even when there was not any“visible” movement. This argues that even though you cannot hear the “why”moving, it really is moving and that is why it is obeying movement constraints.But this movement is “covert,” occurring in the mapping from surface structureto LF, hence not contributing to pronunciation.

The theory with the architecture in (32) was developed in great detail inChomsky (1981a) and came to be known as the Government–Binding (GB)theory.9 In this theory, deep structure, by now generally called “D-structure,”10

has lost its significance as an interface with semantics. It is simply the level thatbegins the syntactic derivation. And surface structure, now called S-structure,is the “branch-point” in the derivation, leading, on one branch, towards PF,the phonetic interface, and on the other, towards LF, the semantic interface.However, it was acknowledged in Chomsky (1981a) that there are semanticphenomena, principally involving anaphoric relations, that seem to require cru-cial reference to surface structure. One such phenomenon concerns anaphoricconnection between pronouns and full NPs. Consider the following pair:


78 Howard Lasnik

(39) John thinks he won.

(40) He thinks John won.

In (39), he can be (though need not be) used to denote John: John and he canbe “coreferential.” But in (40), He cannot be used to denote John: John andHe are necessarily “non-coreferential” in this instance. Coreference betweentwo NPs is standardly represented by coindexing them (that is, by giving themthe same numerical subscript). Thus, we have the grammatical (41) vs. theungrammatical (42):

(41) John1 thinks he1 won.

(42) *He1 thinks John1 won.

The descriptive generalization involves the structural relation “c(onstituent)-command,” which is itself based on “inclusion.” “Inclusion” is the relation ina given structure between, for example, S, on the one hand, and NP and VP,on the other. That is, S includes both NP and VP. Based on these notions, thestructural relation c-command is defined as follows:

(43) In a given structure, a category X c-commands another categoryY if every category which includes X also includes Y.

In particular, one NP c-commands another if every category which includes theformer NP also includes the latter NP. The condition proscribing coreferencein (40) is stated in (45), based on definition (44).11

(44) One NP binds another NP if the former c-commands the latter andthe two NPs are coindexed (have the same index).

(45) A pronoun may not bind a full NP.

Consider (39), in light of definition (44). If John and he are to be coreferential,they will have to be coindexed. Then John binds he in (41), since John bothc-commands he and is coindexed with it. Note that this is allowed by (45), sincethe pronoun does not bind the full NP. If, on the other hand, the two NPs werereversed, as in (42), the pronoun binds the full NP. (45) thus accounts for theungrammaticality of that example (on the specified interpretation).

What Chomsky (1981a) observed is that while overt movement can obvi-ate potential violations of (45), corresponding hypothesized covert movementcannot. Consider the following example, in which the indicated coreference isblocked by (45):

(46) *He1 likes everyone that John1 knows.



Many dialects of English have a process of “Topicalization” by which an NP canbe moved to the front of the sentence. Applying this movement transformationto the direct object in (46) yields the following:

(47) Everyone that John1 knows, he1 likes.

Significantly, coreference between John and he is now possible. So far, thisis as expected, given the basic GB model, since the LF of (47) is presumablyindistinguishable from its surface structure in relevant respects, and in thatsurface structure, shown in (48), he no longer c-commands John as it did indeep structure. This can be seen in the phrase structure tree for (47), where thefronted NP is assumed to be adjoined to the original S:

(48) S

NP S

Everyone that John knows

NP VP

he V NP

likes t

The problem is that the application of QR to (46) yields an LF indistinguish-able from (48), the surface structure and LF of (47). Incorrectly, (46) is pre-dicted to allow coreference, if LF is the sole input to semantic interpretation. InChomsky’s next development in his theory, the minimalist program, this prob-lem becomes acute, as will be seen below.

Levels in a minimalist model

Given the diminishing role of deep and surface structure in the theory, Chomskybegan to consider the possibility that neither is actually a level of representa-tion. Chomsky (1993a) advances a “minimalist program” for linguistics, whichcarries still further the successive simplifications in the theory from the 1950sthrough the 1980s. He argues that if a language is to relate sound to meaning atall, it requires the “interface” levels of LF and PF, the former interfacing with theconceptual-intentional system of the mind, and the latter with the articulatory-perceptual system. Neither D-structure nor S-structure is conceptually neces-sary in this way.12 This motivates a shift to a model that is reminiscent ofChomsky’s original one in the 1950s, with structure building done by general-ized transformations. The derivation begins with a “numeration,” a set of lexicalitems selected from the lexicon. The lexical items are inserted “on-line” in the


80 Howard Lasnik

course of the syntactic derivation. The derivation proceeds “bottom-up” with themost deeply embedded structural unit created first, then combined with anotherlexical item to create a larger phrasal unit, and so on. Furthermore, singularytransformational operations are interspersed with these generalized transforma-tional operations, again roughly as in the much earlier model. Recall, though,that Chomsky (1965) presented a powerful argument against such a model, thatit allowed derivations that never actually occur in human languages. The modelwith recursion in the base excluded those unwanted derivations. However, oncloser inspection, it was not actually elimination of generalized transforma-tions that had this limiting effect. Rather, it was the stipulation that transforma-tions operate strictly cyclically, starting on the most deeply embedded clauseand proceeding monotonically up the tree. Chomsky (1993a) observed that acondition with the same effect can be imposed on the operation of general-ized transformations and their interaction with singulary transformations. Thiscondition, often called the “extension condition,” simply requires that a trans-formational operation “extends” the tree upwards. This guarantees the samesort of monotonic derivations as those permitted by Chomsky (1965). The oneremaining Aspects argument against generalized transformations can also bestraightforwardly addressed. Chomsky had argued that eliminating generalizedtransformations yields a simplified theory, with one class of complex operationsjettisoned in favor of an expanded role for a component that was independentlynecessary, the phrase structure rule component. Further, that simplification wasa substantial step towards answering the fundamental question of how the childselects the correct grammar from a seemingly bewildering array of choices.Eliminating one large class of transformations, generalized transformations,was a step towards addressing this puzzle. This was a very good argument. Butsince then, numerous discoveries and analyses have indicated that the trans-formational component can be dramatically restricted in its descriptive power.In place of the virtually unlimited number of available, highly specific, trans-formations of the theories of the 1950s and early 1960s, we can have insteada tiny number of very general operations: Merge (the generalized transforma-tion, expanded in its role so that it creates even simple clausal structures), Move,Delete. The complex apparent results come not from complex transformations,but from the interactions of very simple ones with each other, and with verygeneral constraints on the operation of transformations and on the ultimatederived outputs. The 1965 argument can then be reversed on itself: eliminatephrase structure rules! This model, similar in significant respects to the originalone in the 1950s, can be graphically represented as in the following diagram,where the point of “spell-out” is where the derivation splits off on one branchtowards PF, ultimately phonetics, while the transformational derivation itself(the “syntactic” portion of the derivation) continues on towards LF, ultimatelysemantics:



(49) Numeration (the selection of lexical items)

Generalized and Singulary Transformations (combining members of the numeration and altering the resulting structures)

Point of “spell-out” PF

LF

A major technical goal of the minimalist program is to reduce all constraintson representation to “bare output conditions,” determined by the propertiesof the systems external to the language faculty (but still internal to the mind)that PF and LF must interface with. Internal to the computational system, thedesideratum is that constraints on transformational derivations will be reducedto general principles of economy. Derivations beginning from the same lexicalchoices (the numeration, in Chomsky’s term) are compared in terms of numberof steps, length of movements, etc., with the less economical ones being rejected.Lexical items are assumed to be composed of “features,” some of which needto be “checked” in particular configurations. This is what drives movement,since, all else being equal, a derivation with an instance of movement is lesseconomical than one without. Further, moving less material is more economicalthan moving more material. This latter point provides a direction for resolvingthe near paradox (under minimalism) that covert operations do not seem toaffect anaphoric possibilities, as illustrated by (46) above. Suppose movementis always driven by the need for some feature F to be checked. Then, as Chomsky(1995a) puts it,13

The operation Move, we now assume, seeks to raise just F. Whatever “extra baggage”is required for convergence involves a kind of “generalized pied-piping.” In an optimaltheory, nothing more should be said about the matter; bare output conditions shoulddetermine just what is carried along, if anything, when F is raised.

For the most part – perhaps completely – it is properties of the phonological componentthat require pied-piping. Isolated features and other scattered parts of words may not besubject to its rules, in which case the derivation is canceled . . . (1995a: 262–3)

Thus, overt movement will almost invariably be of a whole word or phrase.Covert movement, on the other hand, does not feed into phonology, so just thecrucial features will move, leaving the larger category behind. By this line ofreasoning, in the case of (46), John will remain in the c-command domain ofhe at LF, correctly resulting in non-coreference.


82 Howard Lasnik

It is interesting to observe that while the several stages of the theory have allagreed that surface structure (or its minimalist descendant “spell-out”) is thesole input to phonological interpretation, there have been quite distinct claimsabout the connection to semantics. The successive models posited the followingrepresentations as the interface:

(50) a. LSLT model T-markerb. Aspects model Deep Structure

(the “standard theory”)c. Extended Standard Theory Deep and Surface Structured. Government–Binding (GB) LF (via S-structure)e. Minimalism LF (via a continuous

transformational derivationbeginning with the numeration)

Note that in all of these models except the first, we find one (or two) spe-cific syntactic structures that feed into semantic interpretation. In the LSLTmodel, on the other hand, the interface is the T-marker, which includes all ofthe syntactic structures created in the course of the derivation. Now recall thatthe minimalist approach to structure building is much more similar to that inLSLT than to any of the intervening models. This suggests that interpretationin the minimalist model also could be more like that in the LSLT model, dis-tributed over many structures. Interestingly, already in the late 1960s and early1970s, there were occasional arguments for such a model of interpretation evenwithin the Extended Standard Theory, and for phonological interpretation aswell as semantic interpretation. For example, Bresnan (1971) argued that thephonological rule responsible for assigning English sentences their intonationcontour applies cyclically, following each cycle of transformations, rather thanat the end of the entire syntactic derivation. There were similar proposals forsemantic phenomena involving scope and anaphora put forward by Jackend-off (1969).14 In his work of the very late 1990s,15 Chomsky suggests a moregeneral instantiation of this distributed approach to phonological and seman-tic interpretation, based on ideas of Epstein (1999) and Uriagereka (1999).At the end of each cycle (or “phase” in Chomsky’s most recent work), thesyntactic structure thus far created is encapsulated and sent off to the inter-face components for all phonological and semantic interpretation. Thus, whilethere are still what might be called PF and LF components, there are nolevels of PF and LF. Epstein argues that such a move represents a concep-tual simplification, and both Uriagereka and Chomsky provide some empiricaljustification.

The simplifying developments in the theory leading towards the minimalistapproach have generally led to greater breadth and depth of understanding of



both how human languages work and how they are acquired by children. Thissuccess has led Chomsky to put forward the audacious proposal that the humanlanguage faculty might be a “perfect” solution to the problem of relating soundand meaning, given the boundary conditions provided by other modules of themind. Much further research is, of course, needed to determine whether thisbold proposal is sustainable.


4 How the brain begets language

Laura-Ann Petitto

I first met Noam Chomsky through a project that attempted to get the babychimp Nim Chimpsky to “talk.” At nineteen, with the certainty of youth, Iknew that I would soon be “talking to the animals.” Nim was the focus ofour Columbia University research team’s Grand Experiment: could we teachhuman language to other animals through environmental input alone with directinstruction and reinforcement principles? Or would there prove to be aspects ofhuman language that resisted instruction, suggesting that language is a cognitivecapacity that is uniquely human and likely under biological control? Nim wasaffectionately named “Chimpsky” because we were testing some of Chomsky’snativist views. To do so, we used natural sign language. Chimps cannot literallyspeak and cannot learn spoken language. But chimps have hands, arms, andfaces and thus can, in principle, learn the silent language of Deaf people.

By the early 1970s, a surprising number of researchers had turned to learningabout human language through the study of non-human apes. Noam Chomskyhad stated the challenge: important parts of the grammar of human language areinnate and specific to human beings alone. Key among these parts is the specificway that humans arrange words in a sentence (syntax), the ways that humanschange the meanings of words by adding and taking away small meaningfulparts to word stems (morphology), and the ways that a small set of meaninglesssounds are arranged to produce all the words in an entire language (phonology).The human baby, Chomsky argued, is not born a “blank slate” with only thecapacity to learn from direct instruction the sentences that its mother rein-forces in the child’s environment, as had been one of the prevailing tenets of afamous psychologist of the time, B. F. Skinner. Nor are babies born with innateknowledge of a specific language, which had been one caricature of Chomsky’sinnateness views of the time. What is innate in the baby, instead, is tacit knowl-edge of the finite set of possible grammars that world languages could assume(the finite set of units and the relations among them that make up a sententialstring, and the finite ways that they move to form different arrangements insentences). Innately equipped with this tacit knowledge of the finite set of pos-sible language units and the rules for combining them, the baby listens to thepatterns present in the specific language sample to which she is being exposed,

84


How the brain begets language 85

and “chooses” from her innate set of possible grammars the grammar she ishearing. Chomsky’s brilliant theoretical proposals on these topics had so cap-tured the imagination of the international public – and we were all so much inthe thick of arguing for or against the innateness of language and other formsof higher human cognition – that history would soon come to call this periodthe “Chomskyan Revolution.”

My departure from Project Nim Chimpsky in the mid 1970s to attend graduateschool in theoretical linguistics at the University of California, San Diego,was bittersweet. It had become clear that while Nim had some impressivecommunicative and cognitive abilities, there was a fundamental divide betweenhis knowledge and use of language and ours. No one can “talk to the animals”by sign or otherwise. Nim’s data, along with our close analyses of data from allother chimp language projects, unequivocally demonstrated that Chomsky wascorrect: aspects of human language are innate and unique, requiring a humanbiological endowment.

Guided (and inspired) by Noam Chomsky’s theoretical formulations ofhuman syntax and morphology, we discovered that chimpanzee and humansyntax are fundamentally different. While apes can string one or two “words”together in ways that seem patterned, they cannot construct patterned sequencesof three, four, and beyond (“words” and “signs” are homologous). After pro-ducing a “matrix” two words, they then – choosing from only the top five or somost frequently used words that they can produce (all primary food or contactwords, such as eat or tickle) – randomly construct a grocery list. There is norhyme or reason to the list, only a word salad lacking internal organization.Remarkably, moreover, chimps never produce word morphology. They do notseem to have any understanding of a basic word stem, nor of modifying itsmeanings by adding small meaningful word parts (“morphemes”) that we bindor “affix” in highly patterned ways to word stems. If they were to naturallyacquire the word fruit (which they don’t) they would not readily acquire fruity,fruitful, unfruitful, fruitfulness . . . Born with no capacity at all to make thestem/affix distinction, they never – unlike human children, who quite quicklydevelop the ability to understand and use affixed terms – develop it later.

Add to this picture the fact that the actual physical forms of chimp lexicalproductions vary from one time to another in very unsystematic ways. Thisis not a matter of chimps having bad or immature “pronunciation” of theirlexicon, nor is it due to differences between the hands of chimps and humans.Instead, their lexical productions are not patterned and their production errorsare random – not drawn from the finite set of units from which all of their wordsand sentences are built. This fact never changes over chimp development. Inshort, chimps lack sign phonology. It has always interested me that despite thecontroversial abilities attributed to chimpanzees in the “Ape-Language Wars”over the decades, no researcher has ever dared to claim that any chimp has


86 Laura-Ann Petitto

mastered the phonological aspect of human language organization. This deeplytelling fact is returned to below.

Alas, the whole story is even worse than irregularities in chimpanzees’ syntax,morphology, and phonology: the very meanings of their words were “off.” Forone thing, chimps cannot, without great difficulty, acquire the word fruit. Whileapes seem to have some capacity to associate words with concrete things andevents in the world they inhabit, unlike humans, they seem to have little capacityto acquire and readily apply words with an abstract sense. Thus, while chimpscan associate a small set of labels with concrete objects in the world (apple forapples, orange for oranges), they have enormous difficulty acquiring a wordlike fruit, which is a classification of both apples and oranges. There is notangible item in the world that is literally fruit, only instances or examples ofthis abstract kind-concept that seems to exist only in human heads.

For another thing, chimps do not use words in the way we do at all. When wehumans use the common noun apple in reference to that small round and juicyobject in the world that we eat, we do not use it to index (pick out) only oneobject in the world (say, a specific red apple on a table), nor do we use it to referto all things, locations, and actions globally associated with apples. Instead weuse the label to “stand for” or symbolize the set of related objects in the worldthat are true of this particular kind-concept in our heads. Crucially, we also knowthe range or scope over which word kind-concepts may apply: for example, thelabel apple symbolizes a set of related objects and therefore this label is usedonly in reference to objects, not actions. (We further know how kind-conceptssuch as apple act in a sentence, i.e. what forms it can accept, like the noun pluralmarker -s, and what forms it cannot accept, like the verb present progressivemarker -ing.) Although chimps can be experimentally trained to use a labelacross related items (such as the use of the sign apple while in front of a redapple or a green apple), children learn this effortlessly without explicit training,and chimps’ spontaneous label-usage respects none of the above underlyingconstraints. Chimps, unlike humans, use such labels in a way that seems to relyheavily on some global notion of association. A chimp will use the same labelapple to refer to the action of eating apples, the location where apples are kept,events and locations of objects other than apples that happened to be stored withan apple (the knife used to cut it), and so on and so forth – all simultaneously,and without apparent recognition of the relevant differences or the advantagesof being able to distinguish among them. Even the first words of the younghuman baby are used in a kind-concept constrained way (a way that indicatesthat the child’s usage adheres to “natural kind” boundaries – kinds of events,kinds of actions, kinds of objects, etc.). But the usage of chimps, even after yearsof training and communication with humans, never displays this sensitivity todifferences among natural kinds. Surprisingly, then, chimps do not really have“names for things” at all. They have only a hodge-podge of loose associations



with no Chomsky-type internal constraints or categories and rules that governthem (Seidenberg & Petitto 1979, 1987; Terrace et al. 1979). In effect, they donot ever acquire the human word apple.

My disappointment with chimpanzee language was, however, balanced bythe prospect of pursuing an intriguing hypothesis. Because humans can readilyacquire both signed and spoken natural languages, they must, I reasoned, possesssomething at birth in addition to mechanisms for producing and perceivingspeech sounds that makes this possible. I wanted to discover what this elusive“something” could be.

By the mid 1970s, linguistics and psychology (especially adult psycholin-guistics and child developmental psycholinguistics) were abuzz with excitementover Chomsky’s “language acquisition device” (LAD). As stated in generalterms above, the LAD assumes innate knowledge of a set of universal andspecifically linguistic elements and relations. Armed with such knowledge, theyoung child can (i) narrow the range of possible grammars consistent with apartial (and often defective) set of sentences (the “primary linguistic data”) and(ii) fix on a theory (a grammar) for the specific native language to which it isexposed. My specific questions focused on mechanisms: if such a LAD exists,precisely how might the human brain embody it? How might innate, specificallylinguistic knowledge of the set of basic elements and relations be encoded inneural tissue? I knew that attempts to understand this would provide a key towhat biologically distinguishes human language (including human minds andbrains) from the communication of other animals.

The biological foundations of human language

It is not surprising that most linguists, along with most of those who thinkabout the biological foundations of language, closely associate language withspeech. For most of us, speech comes early and remains the primary modalityfor linguistic expression. It is a mistake, however, to associate language andspeech too closely. Natural languages must be defined more abstractly, and thescience of language must be able to deal with evidence from other modalities.An excellent reason for thinking this is that signed languages are acquired atthe same rate as verbally expressed ones. They also reflect the same universals(“principles”). So, language must be defined in a way that applies as easily tosign as it does to speech. To someone like me, who is interested in both thedevelopment of language and its neural embodiment, this fact raises intriguingquestions.

A superficial reading of Chomsky’s early and current work – both formal andinformal – might give the impression that he, like many others, closely asso-ciates language and speech. In his formal work, for example, he calls one of the“interfaces” of the language faculty “phonetic form” (PF) [recently: PHON].



Informally, when arguing against philosophers who seem often to think ofthe words of languages as marks on a page, he points out that speech is cer-tainly prior to marks on page or stone, and that words are sounds in the head.But a closer reading – confirmed by many inspiring discussions that I hadwith Chomsky beginning in the early 1980s when I was a doctoral student atHarvard – indicates that he has a much more abstract characterization oflanguage and its acquisition in mind.

Chomsky is also famous for insisting that his formal view of linguistic com-putation is not a view of “real-time” neural processes. Moreover, he stronglyresists the ideologues who want to tell us that mental processes are “nothing but”neural processes. So, those who study his work might get the impression that hedislikes neural and brain evidence. But this impression too is wrong. His basicview is that the neural investigation of language is still in its very early stagesand at the moment the linguist is in a much better position to tell the neuro-physiologist what to look for than the other way around. He welcomes goodstudies and evidence on the matter. One purpose of this chapter is to describewhat is, I hope, some evidence of this sort.

Studies of very early signed-language acquisition offer an especially clearwindow into the biological foundations of all of human language. Spoken andsigned languages utilize different perceptual modalities (sound versus sight),and the motor control of the tongue and hands are subserved by different neuralsubstrates in the brain. Comparative analyses of these languages, then, promiseinsights into the specific neural architecture that determines early human lan-guage acquisition in our species. If, as has been argued, very early humanlanguage acquisition is under the exclusive control of the maturation of themechanisms for speech production and/or speech perception (Locke 2000;MacNeilage & Davis 2000), then spoken and signed languages should beacquired in radically different ways. At the very least, fundamental differenceswould be predicted in the maturational time course and structure of spokenversus signed language acquisition, presumably due to their use of differentneural substrates in the human brain.

I have conducted comparative studies of monolingual hearing children(groups acquiring English, and others French) and monolingual deaf children(acquiring American Sign Language, ASL, or Langue des Signes Quebecoise,LSQ) from ages birth through 48 months. I have also conducted studies ofyoung bilinguals in “typical” contexts, such as babies acquiring French andEnglish. These bilinguals were compared to two extraordinary cases of child-hood bilingualism: bilingual hearing babies acquiring a signed and a spokenlanguage from birth, as well as bilingual hearing babies acquiring two signedlanguages but no spoken language. Further, I have conducted comparative stud-ies of how the human brain processes highly specific aspects of natural languagestructure in profoundly deaf adults processing signed language as compared to



hearing adults processing spoken language, using modern Positron EmissionTomography (PET) brain-scanning technology. The empirical findings from allof these studies are clear. They show surprising similarities in the overall timecourse and structure of early signed and spoken language acquisition as well asin their neural representation in the human brain. Below, I briefly summarizeeach set of key findings and offer a hypothesis about some of the neurologicalmechanisms that permit human language acquisition to begin. Then I suggestsome implications for Chomsky’s view of language.

Milestone data

Monolingual signing versus speaking babies

Deaf children exposed to signed languages from birth acquire these languagesin the same stages and at the same times as hearing children who acquire spokenlanguages. The stages include the “syllabic babbling stage” (6–10 months) aswell as other developments in babbling, including “variegated babbling” (ages10–12 months), “jargon babbling” (ages 12 months and beyond), the “firstword stage” (9–14 months), the “first two-word stage” (17–26 months), and thegrammatical and semantic developments beyond.

Signing and speaking children also exhibit remarkably similar semantic, dis-course, and pragmatic complexity in their development. For example, analysesof young ASL and LSQ children’s social and conversational patterns of lan-guage use over time, as well as their expressions’ conceptual content, categories,and referential scope, demonstrate unequivocally that their language acquisi-tion follows the identical path seen in age-matched hearing children acquiringspoken language (Petitto 2000).

Bilingual hearing babies acquiring a signed and a spoken language

Recent work focuses on two very unusual populations that provide data richwith theoretical implications: hearing children in bilingual, “bimodal” (signing–speaking) homes, and hearing children who are not exposed to spoken languageat all in early life, only to two signed languages. First, the bilingual hearingchildren exposed to both a signed and a spoken language from birth (e.g. oneparent signs and the other parent speaks) demonstrate no preference what-soever for speech, even though they can hear. For example, these speaking–signing bilingual children acquiring French and LSQ produced their first wordin French and their first sign in LSQ at the same time. Indeed, each of thesesigning–speaking children’s languages are acquired on an identical timetable,and this timetable is the same as for other bilingual children acquiring, forexample, French and English from birth; it is even the same, remarkably, as the



timetable for monolingual children! And contrary to fears of confusing chil-dren by exposing them too early to two languages, bilingual children simulta-neously exposed to two languages from birth achieve their linguistic milestoneson the same timetable as monolinguals, revealing no language delay or con-fusion (Petitto et al. 2001; Charron & Petitto 1991; Holowka, Brosseau-Lapre& Petitto 2002; Kovelman & Petitto 2002, 2003; Petitto & Holowka 2002;Petitto & Kovelman 2003; Petitto, Kovelman, & Harasymowycz 2003). Butthe findings from the signing–speaking children provide us with data that haveparticularly clear theoretical implications. If speech per se were neurologicallyprivileged at birth, then these children might have been expected to glean anymorsel of sound that they could get, perhaps even turning from the visuallysigned input. Instead, they acquire both the signed and the spoken languages towhich they are exposed on an identical maturational timetable.

Second, and perhaps even more surprising, are data from a study I wasfortunate enough to undertake of an extraordinary group of bilingual children.Although these children could hear, their profoundly deaf parents had exposedthem exclusively to two signed languages from birth through early childhood,with no spoken language input. For example, in one family, the deaf motherwas from the United States and signed ASL and the deaf father was fromQuebec and signed LSQ. These children achieved all milestones in their twosigned languages on the same timetable as each other and in the identicalmanner observed in all other bilingual and monolingual children (Petitto 2000).Moreover, this same pattern of development was also observed in yet anotherparticularly interesting group of children – a group of hearing monolingualbabies who were exposed exclusively to one signed language (no speech). Here,as above, these children achieved all of the classic language milestones in signlanguage on the same timetable as hearing babies acquiring speech, includingbabbling on their hands, but not vocally because they had never been exposedto speech (Petitto et al. 2001).

Summarizing so far, entirely normal language acquisition occurs in pro-foundly deaf children exposed only to signed languages, hearing bilingualbabies acquiring a signed and a spoken language simultaneously, and, mostremarkably, hearing children without any spoken language input whatsoever,only signed language input. These data clearly provide no support for the pre-vailing hypothesis that normal human language acquisition in all children isdetermined primarily by the maturation of the mechanisms to hear and producespeech. Interestingly, the hearing bilingual babies who were presented at birthwith a tacit choice (speech versus sign) attended equally to these two input sig-nals, showed no preference for speech whatsoever, and achieved every languagemilestone equally and on the same timetable as monolinguals. Moreover, thehearing babies exposed exclusively to signed language exhibited normal lan-guage acquisition (albeit in sign) and did so without the use of the brain’s



auditory and speech perception mechanisms, and without the use of the motormechanisms used for the production of speech.

Structural data

Homologies in signing and speaking babies

Researchers trying to understand the biological roots of human language havenaturally tried to find its “beginning.” The regular onset of vocal babbling –bababa and the other repetitive, syllabic sounds that infants produce – has ledresearchers to conclude that babbling represents the initial manifestation ofhuman language acquisition, or, at least, of language production. Babbling –and, by extension, early language acquisition in our species – has been said to bedetermined by the development of the anatomy of the vocal tract and the neuro-anatomical and neurophysiological mechanisms subserving the motor controlof speech production. In this view, baby babbling is at first a fundamentallymotoric behavior, rather than a linguistic activity. Here, babies learn languageby pairing these motoric forms – through learned associations – with meaning-ful words in the environment (e.g. MacNeilage & Davis 2000). The existenceof babbling has been further used to argue that the human language capacity isexclusively linked neurologically at birth to innate mechanisms for producingspeech in the development of language in a child, or ontogeny (Liberman &Mattingly 1989). It has also been presented as proof that human languageevolved over the period of human phylogenetic development exclusively fromour species’ incremental motoric ability to control the mouth and the jaw mus-cles (Lieberman 2000).

In 1991, my graduate student Paula Marentette and I reported a surprisingdiscovery, the existence of babbling on the hands of profoundly deaf babies(Petitto & Marentette 1991). Through intensive qualitative analyses of the handsof young deaf babies exposed to sign as compared to hearing babies exposedto speech (ages 10 to 14 months), we found a discrete class of hand activity indeaf babies that was structurally identical to vocal babbling observed in hearingbabies. Like vocal babbling, manual babbling possesses (i) a restricted set of“phonetic” units (unique to signed languages) and (ii) syllabic organization. Itis also (iii) used without meaning or reference. This babbling hand activity wasalso different from all babies’ other hand activity, be they deaf or hearing. Itsstructure was particularly distinct from all babies’ communicative gestures, orthe deaf babies’ attempts to produce real signs.

The discovery of babbling in the silent modality of the hands disconfirmedthe view that babbling is neurologically determined wholly by the maturationof the ability to talk. Instead, it confirmed a claim central to Chomsky’s theory:that early language acquisition is governed by tacit knowledge of the abstract



patterning of language that is biologically endowed in the species, and that thisgovernance is so powerful that it will “out” itself by mapping onto the tongueif given the tongue, or the hands if given the hands – all the while preserv-ing linguistic structures across the two modalities. The deep commonalitiesbetween the linguistic patterns expressed on the tongue in hearing children’svocal babbling and those seen on the hands of deaf children’s silent babbling(independent of the tongue) teach us that Chomsky’s prophetic emphasis onlanguage’s core underlying principles and patterns (not the peripheral ability totalk) are the organizing force behind our extraordinary capacity for language.

It is crucial that the Petitto & Marentette (1991) study discovered the existenceof syllabic organization in the deaf babies’ silent hand babbling. Like spokenlanguage, the structural nucleus of the “sign” (identical to the “word”) in signedlanguages is again the syllable. Although the precise quantitative properties ofthis rhythmic activity were not known at the time (see below), in signed lan-guages, the sign-syllable consists of the rhythmic closing and opening (and/orthe rhythmic hold–movement/movement–hold) alternations of the hands/arms.This sign-syllabic organization has been analyzed as being structurally homol-ogous with the closing and opening of the mouth aperture in the production ofconsonant–vowel, CV (closed–open) mouth alternations in spoken language.The convergence of similar syllabic structures unique to babbling, be it on thehands or the tongue, suggested once again that something other than peripheralfactors, such as the mouth and jaw, was driving this fundamentally linguisticbehavior in young humans. Something else was guiding this powerful con-vergence of structure on two radically different modalities. Discovering whatthis was would bring us closer to discovering the underlying brain mechanisms(should they exist) that could make possible Chomsky’s formal proposals aboutearly language acquisition – his LAD.

A key clue about where to look emerged from the study of deaf babies’hand babbling. When they produced hand babbling, their hands seemed tomove with a different rhythm than their other hand movements – those that allbabies make. But was this difference real? Maybe babies exposed to signedlanguages simply used their hands more than babies exposed to speech. So mycolleagues and I conducted a quantitative study of young baby hands usinginnovative technology, called “Optotrak”— optoelectronic position-tracking –in an attempt to identify the quantitative rhythmic properties that underlie allbabies’ hand activity. But to test the strength of our own views, we wantedto put them through the hardest possible test. So we examined the hands oftypical young hearing babies acquiring spoken language and that rare group ofbabies mentioned above: hearing babies exposed only to signed languages frombirth (no speech). Both groups of babies were equal in all respects, except forthe modality of language input. If babbling (and, by extension, early languageacquisition) is determined by the development of the control of the mouth alone,



then both groups of babies’ hand activity should be the same. Alternatively,if babbling is a linguistic activity that reflects babies’ sensitivity to specificpatterns at the heart of human language and their capacity to use them, thenthe two groups of babies’ hand activity should differ. Indeed, as Chomsky hadargued in his LAD, if babies are born with tacit knowledge of the core patternsthat are universal to all languages, even signed languages, then the linguistichypothesis predicts that differences in the form of language input should yielddifferences in the hand activity of the two groups. In biological terms, tacitknowledge was construed as the baby’s sensitivity to specific patterns at theheart of human language – in particular, the rhythmic patterns that bind syllables,the elementary units of language, into baby babbles, and then into words andsentences.

The precise physical properties of babies’ hand activity were measured byplacing tiny light-emitting diodes (LEDs) on their hands. The LEDs transmittedlight impulses to cameras that, in turn, sent signals into the Optotrak system.This information was then fed into computer software that provided us withthe timing, rate, path movement, velocity, frequency, and sophisticated 3-Dgraphic displays of all baby hand activity. Optotrak computations were calcu-lated “blind” to videotape reference to the babies’ hands (we did not see thebabies’ hands in the first part of the study, only the lighted dots on the computerscreen). Independently, on-line videotapes were made of all babies for post-Optotrak analyses. This method, then, provided the most accurate and rigorousquantitative analysis of moving hands to date, and an advance over previoussubjective classification of baby hands from videotapes.

The quantitative Optotrak analyses revealed that hearing sign-exposed babiesproduced two types of hand activity, while the hearing speech-exposed babiesonly produced one. Sign-exposed babies produced a significantly different typeof low-frequency rhythmical hand activity, with a frequency around 1 Hertz,and another type of high-frequency rhythmical hand activity, with a frequencyaround 2.5–3 Hertz – the type that the speech-exposed babies used nearlyexclusively! Further, sign-exposed babies’ low-frequency hand activity corre-sponded to the rhythmical patterning of adult sign-syllables and, after liftingthe “blind,” videotape data revealed that this hand activity alone exhibited thequalitative properties of silent linguistic hand babbling.1

Remarkably, a dramatic dissociation of two hand-movement types (linguisticvs. motoric) was carved onto a single manual modality differentiated by dif-ferent rhythmical frequencies. This could only occur if babies find salient, andcan make use of, the rhythmical patterning underlying human language. Thisevidence indicates that specific rhythmical patterns underlie baby babbling, andthese reflect highly specific rhythmical sensitivities that babies must be bornwith. These sensitivities correspond to highly specific aspects of the patterningof natural language and almost certainly constitute one of the central biological



mechanisms by which babies discover the patterns of their native grammar inthe linguistic stream around them (Petitto et al. 2001; Petitto et al. 2004).

In a new twist on a classic theme, we further wondered just how similarvery early language perception is across sign and spoken languages. By around4 months, all babies have the universal capacity to discriminate categori-cally all the phonetic-syllabic units found in the world’s spoken languages(such as [ba] and [pa]), even those that they have never heard. But by around14 months, most babies have lost this universal capacity and have instead gainedan increased sensitivity to detect the phonetic contrasts in their native lan-guage. In order to test the neural basis of this capacity – is it a general acousticor a specific linguistic capacity? – we built an infant-controlled HabituationLaboratory and showed hearing monolingual babies (never exposed to sign)moving images of hands. But these hands were phonetic-syllabic units in ASL(below). As with speech, we found that these babies demonstrated categoricaldiscrimination of ASL hand phonetic-syllabic units at age 4 months, which theylost by 14 months (Baker, Isardi, Golinkoff & Petitto 2003; Baker, Sootsman,Golinkoff & Petitto 2003). Intriguing results were then seen in hearing bilin-gual babies (never exposed to sign) who looked like our young monolingualsat age 4 months. But at age 14 months they showed a linguistic “advantage”:they demonstrated increased sensitivity to phonetic units over their monolin-gual peers, suggesting that experience with multiple languages can serve as a“perceptual wedge,” keeping open longer the capacity to discriminate a widerrange of phonetic units than their monolingual peers (Norton, Baker & Petitto2003). Here, from the new perspective of infant language perception, theseresults provide compelling support that the sensitivity to phonetic-syllabic con-trasts is a fundamentally linguistic (not general acoustic) process and part ofthe baby’s biological endowment.

Before closing this line of studies, we decided to take one last look. Thistime we examined only the everyday hearing baby learning a spoken language.Again we asked, “is babbling a linguistic versus a motoric activity?” But nowwe also wanted to understand when the human language capacity emerges inearly life and – crucially, for me, because of a research desire that was bornafter my work with Nim – to find what its neural basis is. Our challenge wasanswering these questions in a way that would not hurt or unsettle young babies.

To gain another perspective on these issues, we carried out another study. Itis a notable fact that adults tend to talk out of the right side of their mouths. Thisseems to be due to the fact that our brain’s left hemisphere is doing the lion’sshare of our language-processing. (We do not see this right mouth asymmetrywhen we speak to others because our brains correct the uneven image.) Intriguedby this fact of cerebral organization, we wondered whether baby babbling mightbe produced more out of the right side of infants’ mouths, thereby reflect-ing the involvement of their left hemisphere’s language-processing centers.



Encouraged by the availability of a non-invasive measure (below) to assess thelaterality of mouth movements in adults, we applied it for the first time to babymouths. This also gave us an opportunity to find out whether there was evidenceof laterality in other forms of mouth activity. Specifically, would babies producesmiles out of the left sides of their mouths (reflecting the involvement of theirright hemisphere’s emotion-processing centers)? And would babies producenon-babbling vocalizations somewhere in between the left and right sides oftheir mouths?

Ten babies were studied at the onset of their babbling stage, five Englishbabies and five French. This was an important study design consideration toensure that no language-specific effects were being revealed on babies’ mouths.The standard measure of mouth laterality – called the “Laterality Index” – wasused, which has been used around the world in the study of adults, especiallyadults after suffering a neuropsychological trauma (e.g. from a stroke) to deter-mine what brain tissue had been impaired and spared. We found that babiesbabbled out of the right side of their mouths, smiled out of the left, and producednon-babbling vocalizations somewhere in between. This study was the first todemonstrate left-hemisphere cerebral specialization for babies while babblingwhich, in turn, suggests that language functions in humans are lateralized froma very early point in development (Holowka & Petitto 2002).2

Summary: significance of studies of early signed and spokenlanguage acquisition

Summarizing these studies of sign–speech homologies, it seems clear thatdespite modality differences, signed and spoken languages are acquired in vir-tually identical ways. The differences observed between children acquiring asigned language versus children acquiring a spoken language are no greaterthan the differences observed between hearing children learning one spokenlanguage, say, Italian, versus another, say, Finnish. These findings cast seriousdoubt on the core hypothesis in very early spoken language acquisition thatthe maturation of mechanisms for the production and/or perception of speechexclusively determine the time course and structure of early human languageacquisition. They also challenge the hypothesis that speech (sound) is critical tonormal language acquisition, and the related hypothesis that speech is uniquelysuited to the brain’s maturational needs in language ontogeny. What these datasuggest, as Chomsky had hypothesized, is that language does indeed have innatecomputational systems. But here is the added observation that language willco-opt whatever provides it with an opportunity to develop in accordance withits innate agenda. This innate agenda seems perfectly happy to accept and uselanguage on the hands (if presented with signed language) or the tongue (ifthe ambient language is a spoken one). That this should be true is stunning



testimony to the power of this innate agenda, that is, the brain’s specified neuralsensitivity to a core set of patterns underlying human language, and the corol-lary fact that an innate agenda must exist in the first place. But, again, where isthis “patterning” taking place?

Testing hypotheses about the biological–neurological foundationsof language with PET studies of signing and speaking adults

The left hemisphere of the human brain has for over one hundred years beenunderstood to be the primary site of language processing (Wernicke 1874).As in early language acquisition, the fundamental explanation for this fact hasbeen that language functions processed at specific left-hemisphere sites reflectits dedication to the motor articulation of speaking or the sensory processingof hearing speech and sound. Contemporary functional imaging studies of thebrain have provided powerful support for this view, including those demonstrat-ing increased regional cerebral blood flow (rCBF) in specific portions of theleft hemisphere when searching, retrieving, and generating information aboutspoken words (specifically, in the left inferior frontal cortex, called the LIFC).This view is especially evident regarding the left Planum Temporale (PT), andto a lesser extent the right PT, which participates in the processing of the mean-ingless phonetic-syllabic units in all spoken language. The left PT forms partof the classically defined Wernicke’s receptive language area, receiving projec-tions from the primary auditory afferent system, and is considered to constitutea unimodal secondary auditory cortex (for a complete report of these issuesand the present PET study under discussion, see Petitto et al. 2000). Thesedata and studies do not, however, resolve the fundamental question of whetherthese brain sites involved in language processing are devoted to speaking andhearing, or whether they constitute tissue that is better thought of as dedicatedto aspects of the patterning of natural language. For these data and studiesdo not exclude the possibility that areas of the brain thought to be “devoted”to speech perception and production are also those employed in human signlanguages.

The existence of natural signed languages provides key insights into whetherlanguage processed at specific brain sites is due to the tissue’s sensitivity tosound per se, or to the patterns encoded within it. In a study to test this, wemeasured rCBF while deaf signers underwent PET brain scans, which we co-registered with their MRI anatomical brain scans. Vital to this study’s designwas our examination of two highly specific levels of language organization insigned languages, including the generation of signs (lexical level) and phonetic-syllabic units (sublexical level; meaningless parts of signs). As I mentionedbefore, this level of language organization is found in all the world’s languages(be they signed or spoken) and comprises the restricted set of meaningless



units from which a particular natural language is constructed.3 If the brain sitesunderlying the processing of words and parts of words are specialized specifi-cally for sound, then deaf people’s processing of signs and parts of signs shouldengage cerebral tissue different from that classically linked to speech. Con-versely, if the human brain possesses sensitivity to aspects of the patterning ofnatural language, then deaf signers processing these specific levels of languageorganization may engage tissue similar to that observed in hearing speakers.

We studied two entirely distinct cultural groups of deaf people who usedtwo distinct natural signed languages. Five were native adult signers of Amer-ican Sign Language (ASL; used in the United States and parts of Canada) andsix were native adult signers of Langue des Signes Quebecoise (LSQ; used inQuebec and other parts of French Canada). ASL and LSQ are grammaticallyautonomous signed languages; our use of two distinct signed languages consti-tutes another significant design consideration, unique to the present research,introduced to provide independent, crosslinguistic replication of the findingswithin a single study. We further compared these eleven deaf people to tenEnglish-speaking hearing adult controls who had no knowledge of signedlanguages.

Two main findings emerged from this Petitto et al. (2000) study. First, boththe deaf people processing genuine signs and the hearing controls processingwords exhibited clear cerebral blood flow increases within the identical brainregion, the left inferior frontal cortex (LIFC). This finding demonstrated thatone component of processing human language (something as abstract as lexicalsearch and retrieval) was housed at a specific brain site. Because both languageon the hands and language on the tongue were processed at the same brainsite, it supported a surprising hypothesis: there exists tissue in the human braindedicated to a function of human language structure independent of speech orsound.

The second major discovery involved tissue universally viewed as beingliterally tied to sound processing, again the Planum Temporale or PT, especiallythe processing of the small phonetic-syllabic units that make up a spoken word.Here we witnessed robust activation in the profoundly deaf people’s PT whilethey were processing meaningless parts of signs (phonetic-syllabic or sublexicalparts of a sign on the hands). This was the remarkable thing: how could therebe activity in sound tissue in the brains of profoundly deaf people who neverheard sound? The activity could not be due to processing based on any auditoryrepresentations as they are traditionally understood – the transduction of soundwaves, and their pressure on the inner ear, into neural signals. In witnessingthese specific results, we demonstrated neural activity in what has hitherto beenthought to be exclusively auditory cortex by using purely visual stimuli – but,crucially, the visual stimuli were linguistic. Thus, rather than being dedicatedexclusively to sound (as had been thought for generations) it must be that this



tissue is instead dedicated to linguistic patterns in the input – specifically tothe patterns inherent in rhythmically contrasting phonetic-syllabic units – bethey patterns on the hands or the tongue. In short, we found the biologicalinstantiation of a key level of Chomsky’s “hierarchical” levels of languageorganization: phonology.

Finally, we randomized the MRI anatomical brain scans of all of these deafand hearing subjects, flipped the x-axis so that no one knew if they were lookingat the subjects’ left or right hemispheres, and then computed the gray- and white-matter tissue volumes in all of these brains’ classic sound tissue (primary andsecondary auditory cortices) – without ever knowing the hearing status of thebrains being analyzed. In a nutshell, we found that there were no differences inthe gray-matter volumes in the deaf and hearing peoples’ sound tissue (meaningthat there was no cell loss in the sound tissue of profoundly deaf people ascompared to hearing people) and no differences in the white-matter volumesbetween the groups (meaning that there was no loss of neuronal input to thissound tissue). Surprisingly, like hearing people, there was a greater left-versus-right-hemisphere asymmetry in the sound tissue of the deaf people (for a fullreport of these findings see Penhune et al. 2003). How could this be? Whydoesn’t sound tissue shrivel up and die in deaf brains? Here, as above, it mustbe that such tissue is sensitive to specific linguistic patterns in natural language(not sound) and the on-going processing of sign language provides the tissuewith just those linguistic patterns to keep it alive and kicking!

Together such facts demand that those who study language and its acquisitionintroduce hypotheses of the mind/brain that make sense of how this is possible.

Adaptive phonological differentiation

When examining the sublexical level of language organization of signedand spoken languages we find striking commonalities: both employ a highlyrestricted set of units organized into regular structured patterns – patterns thatamount to rapid rhythmically alternating maximal contrasts. This suggests anhypothesis (albeit in a nascent form) that speaks to how visual images mightactivate auditory brain tissue. And it might at least focus further research efforts,for example, to explain what exactly it is about the neurons of the human PT andtheir connections to other systems in the head that gives the specific multimodallinguistic-pattern-responding character it has. The PT can be activated eitherby sight or sound because, I suggest, this tissue (or at least a part of it) hasspecific neurons or groups of neurons working in concert that, when activatedby appropriate input patterns, responds selectively to specific distributions ofcomplex, low-level units in rapid rhythmic alternation. These distributions arethose that, informally, we think of as natural-language phonological structure.To be sure, PT tissue does also, in general, deal with sensory sound input – the



PT is typically employed in hearing humans in processing non-linguistic soundinputs, and its homologue in the brains of some apes apparently performs onlythis task. But, due to some yet unknown factor, PT tissue in humans has a sensi-tivity to certain specific patterns found only in natural languages. This actuallymay be one key neural difference between the chimpanzee and human brain andprovides an intriguing experiment in nature regarding how far a creature couldget in language without it: chimps can hear speech but have no brain power tofind in the stream of sounds around them the finite set of units and their patternsthat make up a language’s phonological inventory. From this, we can predict justabout how far they’d get without this capacity: no syntax, no morphology, and,of course, no phonology. But, with decent memory and association powers,they would be able to pick out or refer to things in their here and now withlist-like global association. Voila. This is just about what chimps do.

Crucially, the PT is apparently not neurally sensitive to any and all rhyth-mically alternating acoustic input containing contrasts. Music, for example,provides complex multifaceted rhythmical signals that engage brain tissue atmultiple cerebral sites yet, in general, contemporary scientists agree that thePT (especially the left PT) is not the brain site for processing these differentforms of rhythmically alternating contrasts (for a review see Zatorre & Binder2000).4 To summarize, the hypothesis is that the left PT site contains, in additionto other forms of specialization, specialization for highly specific, maximallycontrasting rhythmical patterns in the input. These patterns are found exclu-sively in specific aspects of natural language – specifically, phonetic-syllabicunits and their distributional patterning.

If this is correct, there is an initial biologically guided capacity (what I calledthe innate agenda, above) to find salient, and to attend to particular aspects of,input streams involving phonetic-syllabic contrasting units, which after severalmonths of life can (given relevant input) become attuned to whatever sorts ofsensory input are capable – consistent with the internal agenda – of activatingit. Thus, experience with specific phonetic-syllabic units in the input streamprovided by – in the case of humans – sound or vision to a young baby lit-erally changes and “adapts” the baby’s biological perceptual and attentionalmechanisms to be sensitive to patterns in the given modality or modalities.This “guiding” capacity amounts to a neurally set agenda (a system that “looksfor” certain patterns) and leads to the child’s ability to discover, and to utilize,elementary units of language structure. Without such an internal agenda, thechild’s mind would not recognize the patterns needed for linguistic develop-ment. Nim’s mind, lacking what Chomsky calls a “language faculty,” wouldnot and could not find that pattern in the input stream. It would never becomesalient. The PT neurally embodies a small part of the tacit knowledge of the setof basic elements and relations that Chomsky proposed must be contained in achild’s LAD.



On the neural tissue underlying human language acquisition

Returning to the question with which I began my research – how does the brainpermit the radical change in the morphology of its expressive and receptivemechanisms for language found in speech and sign, and what is the geneticbasis for this stunning equipotentiality? – I think we have found at least someanswers. The various studies discussed above suggest that the brain at birth can-not be working under rigid genetic instruction to produce and receive languagevia the auditory-speech modality. If this were the case, then the nature of signedand spoken language acquisition – including the nature of the maturational timecourse and early language structures – as well as the cerebral organization of signand speech in the adult brain should be different. Clearly, it is not. The fact thatthe brain can tolerate variation in language transmission and reception, despitedifferent environmental inputs, and still achieve the target capacity (being aspeaker of a natural language, perhaps several), provides support for a geneticcomponent underlying language acquisition that is nevertheless biologically“flexible” (neurologically plastic). I hypothesized that PT tissue constitutes akey brain site that contributes to launching human language acquisition and havesuggested that it gains its vital role in the establishment of nascent phonologicalrepresentations in all humans through a process that I have termed “AdaptivePhonological Differentiation.” This process “guides” the newborn’s attentionto find salient specific aspects of the input stream with specific rhythmical con-trasts that correspond to key aspects of natural language structure: elementaryphonetic-syllabic units and their sing-song distributional patterning (prosody).Drawing from the baby Optotrak findings mentioned above, I further suggestthat this PT tissue tunes the infant’s perceptual systems to find salient, andto attend to, (initially) maximally-contrasting, rhythmically-oscillating bundlesof about 1.2–1.5 seconds. Armed with this honed sensitivity, the baby’s mindcan, in turn, begin to “select” the restricted set of elementary phonetic unitsand combinatorial regularities of their native language(s). The precise timingis unclear, but it is known that they begin the production of these elementaryunits at around six months (see also Jusczyk 1999).

The same processes must be at work when a baby is confronted at birth withtwo or more natural languages, whether spoken or signed. Here the newborn’ssensitivity to specific rhythmical and distributing patterning must provide itwith the means to detect two related but different rhythmically contrastinglinguistic patterns. The development of this capacity surely serves as a basisupon which bilingual babies tacitly build up representations of their two distinctphonological systems. (Petitto et al. 2001; see also Holowka, Brosseau-Lapre &Petitto 2002; Petitto & Holowka 2002 for a discussion of the processes thatmake possible human bilingual acquisition). Again, exact timing in the case ofmultiple languages is unclear, but this process is certainly well underway by



age 6 months, exhibiting regular growth and expansion in the capacity to detectdistinct forms of systematic rhythmical-temporal and distributional patternsover time. So, whether one language or many, a baby’s innate mechanismswill – irrespective of whether the input is from eye or ear – guide it to findspecific patterns in the input stream and, when its internal systems find them,they “instruct” motor systems to produce “output” informed by them.

Chomsky may not have encountered languages on the hands early in his lifebut, remarkably, his “abstract” theory of the LAD allowed for the flexibility inmodality we have seen in this chapter. And while he assumed the LAD mustbe biologically instantiated in some way, he did not have any idea of how it iswritten into neural tissue. I think we now have some idea of how a small, butcrucial, part of it is.


5 Chomsky and Halle’s revolution in phonology

B. Elan Dresher

Introduction

Chomsky and Halle’s approach to phonological theory, as with other compo-nents of generative grammar, represented a sharp break with the main cur-rents of American linguistics that immediately preceded them. The differ-ences were conceptual as well as technical. Accounts of the development ofphonology emphasize technical issues, such as arguments over the existenceof a “taxonomic phonemic level,” or whether it is permissible to “mix levels”in a phonological analysis. Lying behind discussion of these issues, however,were assumptions about psychology and the practice of science. Indeed,throughout the development of phonology, major changes came about not onlythrough technical breakthroughs, but also by reinterpreting the significance ofexisting technical devices. This was also the case with Chomsky and Halle’sinnovations.

In this chapter I discuss Chomsky and Halle’s contributions to phonologicaltheory by putting their views in the context of the theories that prevailed beforethem. I will also try to connect the technical issues to the larger conceptual onesconcerning the nature of language acquisition and the mind.1 I will be treatingChomsky and Halle’s contributions together, without attempting to distinguishwho contributed precisely which ideas. Their early work in generative phonol-ogy, culminating in the major work The Sound Pattern of English (Chomsky &Halle 1968, henceforth SPE), was done jointly.

Nevertheless, some indication of what each brought to the enterprise can begleaned from Chomsky’s 1957b review of Jakobson and Halle’s Fundamentalsof Language (Jakobson & Halle 1956). Chomsky finds that “much can be said”for Jakobson and Halle’s approach to phonology. In particular, he approved ofthe hypothesis that the sound systems of all languages could be characterizedin terms of a limited number of universal distinctive features. Second, he pre-ferred their approach to identifying phonemes over others then current. Theyassigned two segments to the same phoneme if they have the same feature spec-ifications. Most other approaches to phonemic analysis prevailing at the timeassigned sounds to phonemes if they are in complementary distribution (or in

102


Chomsky and Halle’s revolution in phonology 103

free variation) and phonetically similar, appealing to a notion of similarity thatis difficult to define. Finally, Chomsky seconds the authors’ emphasis (advancedover the years by Jakobson) on the importance of extending phonological theoryto account for language acquisition, disorders, and other aspects of linguisticbehavior.

On the other side, Chomsky observes that many of Jakobson and Halle’sproposals need to be made more explicit and precise before they can be empir-ically tested. He further proposes an amendment to their conception of howphonemes are related to speech. He found their requirement that the distinctivefeatures assigned to phonemes be present in their correct sequence in the pho-netics too strict. He proposes that distinctive feature specifications form insteadan “abstract underlying system of classification related, perhaps indirectly, tothe physical facts of speech.” Finally, Chomsky proposes that general criteriaof simplicity play an important role in the evaluation of particular phonologicalanalyses.

One can say, then, that Chomsky and Halle’s theory of generative phonologywas a synthesis of Jakobson and Halle’s theory of distinctive features and phone-mic analysis, revised in the light of Chomsky’s emphasis on formal explicitness,simplicity, and abstractness and autonomy of mental representations.

Rules and derivations

When first introduced, the centrality of rules in Chomsky and Halle’s approachto phonology appeared revolutionary. A grammar of a language must merely listmany things – for example, the English word tide begins with a t, ends with a d,and has a vowel sound represented by i. A person who knows English but whohappens never to have encountered this word cannot derive this information.It is a particular fact about English that must be learned and committed tomemory.

Other facts about the pronunciation of this word are more systematic. Forexample, the t in tide is pronounced with a puff of air, called aspiration (repre-sented as th), in contrast to the t in style, which is not aspirated. Any speaker ofEnglish told that tide begins with t would automatically know that the t must bepronounced with aspiration. That is, the aspiration of t is not an idiosyncraticfact that must be listed in the lexical entry of tide, but can be encoded in a rule.Thus, the lexical, or underlying, form of the word tide need only specify thatthe initial sound is a /t/, where slant brackets represent phonemic forms; thisform is then subject to the rule of aspiration, which derives the phonetic, orsurface, form [th] (where square brackets represent phonetic forms).

The vowel written i is a diphthong, phonetically [a�j], where � indicatesthat the vowel is long, and j represents a glide. The length of the vowel ispredictable: tide ends in a d, which is a voiced sound, and in English, stressed


104 B. Elan Dresher

vowels lengthen before voiced sounds. Thus, the vowel of bid is longer thanthe vowel of bit, which ends in a voiceless sound, t. Similarly, the diphthongin tight is shorter than in tide. In some dialects, such as Canadian English, thefirst part of this diphthong is pronounced with a higher and more centralizedtongue position before a voiceless consonant, and is phonetically transcribedas [�j]. Therefore, speakers of these English dialects need only learn that thediphthongs in tide and tight are both /aj/. General rules then apply to lengthen/aj/ to [a�j] before voiced sounds and to raise it to [�j] before voiceless sounds.

One might suppose, as it generally was in pre-generative phonology, that thedistribution of the various phonetic realizations, or allophones, of a phonemecould be represented by an unordered set of statements. The diphthong /aj/,for example, appears as [a�j] before voiced sounds (a process we will callLengthening) and as [�j] before voiceless sounds (Raising).

Chomsky and Halle proposed, however, that rules must be ordered if theyare to give correct results and be statable in the simplest, most general way.Consider, for example the words ride and write. They undergo the rules ofLengthening and (in the dialects under consideration) Raising. The rules canbe written as follows:2

(1) Lengthening V → [+long]/—— (glide)

[C

+voiced

]

(2) Raising /a/ → �/—— glide

[C

−voiced

]

These rules also apply in rider and writer. In the pronunciation of NorthAmerican English, the t and d in these words are pronounced with an alve-olar “flap,” in phonetic transcription [ɾ], a quick tap of the tongue rather than asustained occlusion.

(3) Flapping {t,d} → ɾ / V (glide)——

[V

−stressed

]

The result of applying the three rules is shown below.

(4) A simple derivation

writer riderUnderlying /rajtər / /rajdər /Lengthening – ra�jdərRaising r�jtər –Flapping r�jɾər ra�jɾərPhonetic [r�jɾər] [ra�jɾər]

Note that the Flapping rule must follow the other two rules. If Flapping wereto apply first, an incorrect form would be generated: as shown in (5), writerwould be pronounced just like rider, which is not the case in this dialect.3



(5) An incorrect derivation

writer riderUnderlying /rajtər / / rajdər /Flapping rajɾər rajɾərLengthening ra�jdər ra�jdərRaising – –Phonetic *[ra�jɾər] [ ra�jɾər]

Therefore, the basic architecture of the phonological theory of SPE can bediagrammed as in (6).

(6) Basic architecture of phonological component (SPE)4

Underlying forms

(stored in lexicon)

Systematic phonemic level

Set of ordered rules

Surface forms

(closer to pronunciation)

Systematic phonetic level

In hindsight, one might wonder why the rather simple model in (6) wouldhave ever been considered revolutionary. None of the basic ingredients werenovel: not the idea of two basic levels, nor even the idea of a derivation mediatedby ordered rules. However, in the context of phonological theory in Americain the 1950s, it represented a significant new departure. To see why this was sorequires a brief excursion to the nineteenth-century origins of modern phoneticsand phonology.

Two levels: broad and narrow transcription

That at least two levels of representation are required to represent the sounds ofa language was becoming apparent already in the nineteenth century. Phonolo-gists and phoneticians realized that a degree of precision in the representationof sounds was required that was unattainable using conventional alphabets.They aimed to develop a system in which one sound was always representedby one symbol (unlike English, where the sound [s] is sometimes representedby <c> as in <city>, and sometimes by <s>, as in <sit>), and in which


106 B. Elan Dresher

one symbol is used for only one sound (again unlike English, where the letter<c> sometimes represents [s], and sometimes [k] (<electricity>). This move-ment ultimately led to the development of the International Phonetic Alphabet(IPA), a transcription system that approaches the goal of a one-to-one relationbetween sounds and symbols. This type of transcription came to be known asnarrow transcription, because it records very fine distinctions between speechsounds.

It quickly became apparent that a narrow transcription is not a practical wayto transcribe particular languages. For example, consider English t in wordslike stop, top, hat, not you, trap, and writer. When looked at closely, these [t]sare all different: unaspirated [t ] in stop; aspirated [th] in top; unreleased [t−]in hat (optionally: it may also be released and aspirated); palatalized [tj] in notyou; retroflexed [�] in trap; and flapped [ɾ] in writer. These are only some ofthe realizations of English t. Detailed examination of other sounds of Englishreveals that they, too, are not unitary sounds but groups of sounds, distinguish-able by separate phonetic symbols. The IPA has a way of distinguishing allthese sounds, and this is desirable if we wish to give an accurate transcriptionof what each variant actually is. But it would be very cumbersome and quiteimpractical to actually attempt to use this type of transcription as a way ofwriting English.

More important, a narrow transcription fails to do justice to some basicfacts about the sound system of English. For there is something correctabout the intuitions of speakers that the sounds listed above are all “variantsof t.” A transcription system that treats [t] and [th] as being as differentfrom each other as each is to [p] or [n] is missing something important: theEnglish spelling system, for all its faults and quirks, does a better job atcapturing the way sounds actually pattern in English. Thus, alongside nar-row transcription there developed the notion of a broad transcription, whichis designed to abstract away from predictable variations and alternations insounds.

In the above example, [t] and [th] are allophones of the same phoneme /t/,whereas [n] in nip is an allophone of a different phoneme, /n/. We know that /t/and /n/ are different phonemes in English because they are in contrast: tap andnap are different words in English, as are fit and fin. So one function of a broadtranscription is that it abstracts away from allophonic variation and representsonly contrastive differences of sounds.

Rule-governed behavior in sound systems is not limited to allophonic varia-tion. Consider the English plural -s. Following a voiced sound it is pronounced[z], as in dogs, beds, bees, sins, and dolls. Following a voiceless sound, it ispronounced [s], as in cats, ropes, and sticks.5 Since this alternation between[s] and [z] is rule-governed and entirely predictable, it is plausible to supposethat the regular plural morpheme has a single lexical representation, say /z/,



and that English speakers apply a rule devoicing /z/ to [s] following a voicelesssound.

English s and z, however, are not merely allophonic variants of a singlephoneme in English. When these sounds do not immediately follow a consonant,they contrast, as in sip vs. zip, and bus vs. buzz. Therefore, the phonologicalrules in (6) also include rules that change one underlying phoneme into another,in addition to rules that create allophones of a phoneme (without changingphonemic identity). In the case of the plural, the English spelling system againmore closely approximates a broad than a narrow transcription, consistentlywriting the regular plural as <s> even when it is pronounced [z]. Similarly,the final segment in electric is consistently written with a <c>, whether it ispronounced [k], or [s] (as in electricity), or [s] (as in electrician).

Whereas a narrow transcription should ideally be universal, a broad tran-scription is language particular, reflecting the patterning of sounds in particularlanguages.

Narrow transcription in phonological theory

Students of phonology brought up in the tradition of generative grammarwill readily identify broad transcription with Chomsky and Halle’s systematicphonemic level and narrow transcription with the systematic phonetic level.Indeed, the model in (6) appears to be a natural translation into phonologi-cal theory of the two types of transcription. However, some difficulties had tobe overcome in arriving at the model in (6). The first of these concerns thenature of the phonetic level: to what extent is it truly a “systematic” level ofrepresentation?

Leonard Bloomfield (1887–1949), and the American linguists who followedhim, known as the post-Bloomfieldians, maintained that a phonetic level corre-sponding to a narrow transcription cannot be supported as a legitimate linguisticrepresentation because it is not systematic, but arbitrary. According to Bloom-field, such a transcription is dependent on the background and perception ofthe transcriber: some transcribers will notice and note down certain subphone-mic distinctions, but others that are less familiar to them will go unrecorded,particularly as they are not crucial to marking contrastive sounds in thelanguage.

For example, an English-speaking transcriber might record that the t in theEnglish word two is aspirated, because the distribution of aspirated and unaspi-rated /t/ in English is systematic. But there are many other aspects of this soundthat may or may not be noted: whether the sound is dental (made with the tongueagainst the teeth) or alveolar (tongue against the alveolar ridge); whether thelips are rounded (as they are in two) and, if so, how much; whether the tongue isreleased quickly and simultaneously with the puff of air, or whether the tongue


108 B. Elan Dresher

lags a bit, creating an affricated (tending to [ts]) or palatalized [t j] sound; andso on.

Since a linguistic representation must be based on more than just the whimsof individual transcribers, Bloomfield concluded that there is no principled levelof phonetic representation corresponding to a narrow transcription.

As pointed out by Chomsky (1964), this argument rests on the assumptionthat there is no universal theory of phonetic representation. Lacking such atheory, it would appear that a phonetic representation has no principled basis.However, a universal feature theory, of the sort initiated by Prague Schoollinguists and developed in works such as Jakobson, Fant, and Halle (1952),Jakobson and Halle (1956), and subsequently revised by Chomsky and Halle(1968), can serve as the basis for a phonetic transcription. The universal setof distinctive features is designed to discriminate all and only those aspects ofsounds that are contrastive in the languages of the world. SPE, for example,uses twelve distinctive features to represent the consonant sounds of English.6

The existence of a universal set of phonetic features constrains what can go intoa phonetic representation. No such theory existed in American linguistics, sothere was no basis for a systematic phonetic level.

How, then, are the sounds of a language to be represented in a linguisticdescription? We are left with broad transcription, or, in terms of (6), the sys-tematic phonemic level. However, the systematic phonemic level is too remotefrom the surface phonetics, that is, too abstract, to serve as the only level ofphonological representation.

Consider the English vowel system. In English, unstressed vowels tend toreduce to schwa [ə] in many contexts:

(7) English vowel–schwa alternations

Tense vowel Reduced Lax vowel ReducedCanadian [ej] Canada [ə] Asiatic [æ] Asia [ə]managerial [ij] manager [ə] telegraphy [ε] telegraph [ə]horızon [aj] horizontal [ə] medıcinal [i] medicine [ə]custodian [ow] custody [ə] photography [a] photograph [ə]sulfuric [juw] sulfur [ə] production [�] product [ə]

Thus, [ə] is an allophone of every English vowel phoneme. It follows thata phonemic representation of the above words should include unreducedvowels only; reduction to schwa would then be a rule-governed allophonicvariation.

Bloch (1941) argued that while such a system is indeed elegant, it poses prob-lems for a learner (as well as a linguist unfamiliar with the language). What hap-pens when learners come across a schwa whose unreduced version is unknownto them, as in words like sofa or of ? Or even manager, if they haven’t heard a



related form such as managerial? If there were a “lower” phonetic level of rep-resentation, a learner could at least represent the phonetic form of such wordswith a schwa, while deferring a decision as to which underlying phoneme toassign it to. But, having rejected a phonetic level, post-Bloomfieldian theory hadno recourse to such a level of representation. The consequence is that learners(and linguists) would be unable to assign any phonological representation tosuch utterances.

Moreover, according to Bloch, the only data relevant to phonemic analysis are“the facts of pronunciation,” that is, the distribution of surface allophones, andnot, for example, the existence of morphologically related forms. This assump-tion severely limits the evidence one can use in arriving at a phonologicalanalysis. It presupposes an analyst who has no access to the fact that the wordmanager is related to managerial. Such an analyst would not be in a positionto know that the final schwa of the former is related to the stressed vowel of thelatter.

Thus, without a systematic phonetic level, the post-Bloomfieldians neededa new level of representation that was much less abstract than the systematicphonemic level, and that did not suffer from the arbitrariness they attributedto phonetic representation. In structuralist terminology, this level was simplycalled the phonemic representation, and the more abstract systematic phonemiclevel was called the morphophonemic representation. In the terminology ofChomsky (1964), the new level is called the taxonomic phonemic level. Thepost-Bloomfieldian conception of the phonological component was thus asin (8).

(8) Levels in post-Bloomfieldian American structuralist phonology

Underlying forms

stored in lexicon

Morphophonemic level =Systematic phonemic level

Morphophonem

ics


Phonemic forms(Taxonomic) phonemiclevel

Informal statements ofallophonic distribution

Phonetic forms

Phonemics (Not a level of linguistic

theory)


110 B. Elan Dresher

The new (taxonomic) phonemic level

Bloch’s argument assumes that language learners must be able to encode utter-ances into phonemes based only on the distribution of surface sounds, or phones.Of course, learners of a language must acquire not only the phonological system,but the rest of the grammar as well, including the morphology and syntax. How-ever, it became an entrenched assumption of American structuralist linguisticsthat acquisition of language went “bottom-up,” from phones to phonemes, fromphonemes to morphemes, from morphemes to syntax, and so on. Though thisassumption had no empirical support whatsoever, it had important consequencesfor the development of phonological theory.

One consequence was the dictum that it is impermissible to “mix levels” indeveloping a phonemic analysis. That is, a phonemic analysis must be justifi-able solely on the basis of allophonic distribution, making no appeal to “higher”levels such as morpheme identity. As Chomsky (1964) showed, this assumptionhad disastrous consequences for the generality and simplicity of the phonolog-ical analysis.7

An example is Hockett’s (1951) discussion of a hypothetical language with nounderlying contrast between voiced and voiceless consonants: all consonantsare voiceless except word-medially between vowels, where they are voiced.This is a fairly common situation, and the natural assumption is that a phone-mic representation should indicate only voiceless consonants, since voicing ispredictable. Hockett considers the case of two sequences of words in such alanguage, pat adak and padat ak. In the standard analysis, these words wouldhave the phonemic representations shown in (9), where # represents a wordboundary separating the words.

(9) Consonants voiced only word-medially between vowels

a. b.Phonemic / #pat#atak# / / #patat#ak# /Voicing #pat#adak# #padat#ak#Phonetic [patadak] [padatak]

Hockett argues that there is a problem with this analysis. The word boundariesdo not correspond to any sound, or even to a regular absence of sound or pause.If one hears (9b) [padatak] one would not know, on phonetic grounds alone,whether it derives from /patat # ak/ or /pata # tak/. Of course, one could deter-mine this if one knew something about the lexicon: one might find, for example,that there is a stem / patat- / with the appropriate meaning but no stem / pata/.But performing such a look-up is to “mix levels.” In Hockett’s interpretationof phonemic theory, the phonemic representation must not rely on the properpositioning of word boundaries. If boundaries are omitted from the phonemicrepresentation, then both utterances in (9) are represented as / patatak/. But now



we cannot account for the different distribution of voiceless and voiced conso-nants in (9a) and (9b); the only solution, according to Hockett, is to representvoiced consonants as such in phonemic representation. Rather than the phone-mic forms in (9), we would posit / patadak/ for (9a) and / padatak/ for (9b).This result appears to be incompatible with the original concept of a phoneme.Moreover, the generalization that the voicing of consonants is predictable andtherefore need not be learned on a case-by-case basis is lost.8

The attempt to constrain the phonemic level so as to keep it closer to thesurface phonetics led to the requirement that the phonemic level meet a numberof further conditions. Their effect was to ensure that there be a one-to-onerelation (biuniqueness) between allophones and phonemes: given an allophone,it should be possible to unambiguously assign it to a phoneme; and given aphoneme, it must be clear what allophone instantiates it in any given context. The# boundaries in (9) violate biuniqueness, because phonetic [padatak] can deriveeither from /patat# ak/ or /pata# tak/. Like the prohibition on mixing of levels,these conditions resulted in a loss of generalizations, with no compensatorygain in descriptive or explanatory force.

Consider again the interaction of Flapping and rules affecting the /aj/ diph-thong shown above in the writer/rider example (4), repeated here.

(4) A simple derivation

writer riderUnderlying / rajtər / / rajdər /Lengthening – ra�jdərRaising r�jtərFlapping r�jɾər ra�jɾərPhonetic [r�jɾər] [ra�jɾər]

It is clear in (4) that [ɾ] is a predictable allophone of both / t / and / d/, and thereis no difficulty in formulating rules to account for its distribution. However,this simple derivation fails a number of conditions that the post-Bloomfieldianlinguists placed on phonemic representations.

First, the phonemes / t / and / d/ have a common allophone, [ɾ]. This amountsto a partial overlapping of the two phonemes, which violates the biuniquenesscondition. The problem with overlapping is that it is not possible, upon inspec-tion, to decide which phoneme an allophone belongs to. Of course, if we couldappeal to morpheme identity we would know that the [ɾ] of writer belongs to/ t /, because of write, and that the [ɾ] of rider belongs to / d/, because of ride.But this again violates the constraint against mixing of levels.

Another problem with this analysis from the point of view of post-Bloomfieldian theory is that there is a mismatch between the location of the


112 B. Elan Dresher

phonemic and the phonetic contrast in writer and rider: forms that are phonemi-cally different only in their fourth member (/ t / vs. / d/) are phonetically differentonly in their second member ([�j] vs. [a�j]). According to Chomsky (1964), thismismatch is a violation of the condition of linearity.

This example shows also that the notion of minimal pair is not a self-evidentone. A minimal pair is a pair of words that differ in a single phoneme. Minimalpairs are often used to show that two sounds contrast in a language. For example,we can demonstrate that [s] and [z] contrast in English by adducing minimalpairs such as sip and zip, or bus and buzz. Since the only difference in thesewords is the [s] vs. [z], we conclude that they belong to distinct phonemes.However, a similar test would show that [a�j] and [�j] are distinct phonemesin English, since writer and rider appear to be minimal pairs distinguished intheir second elements, not their fourth.

As Chomsky (1964) points out, minimal pairs are thus not evident from thesurface, but require that we take into account various kinds of information. Inthe case of sip and zip there are no further facts that contradict the conclusionthat the distinction is simply between / s / and / z/.9

According to the logic of post-Bloomfieldian phonemics, then, we wouldhave to transcribe writer as / r�jɾər / and rider as / ra�jɾər /. In effect, thephonemic level would fail to capture any of the generalizations about Englishsound patterns discussed above. It would fail to note that [ɾ] is a predictableallophone of / t / and / d/, and that [�j] and [a�j] are predictable allophonesof / aj /.

Finally, we have seen that the rules in (4) have to be ordered; ordering wasnot permissible in American structuralist phonemics. The relationship betweena taxonomic phoneme and its allophones had to be statable as a set of unordereddistributional statements.

What, then, of the generalizations about sound patterning that are therebyexcluded from the phonemics? Where in the grammar, for example, do we rep-resent the fact that there is a single regular English plural, or that the soundsof write are systematically related to the sounds of writer, or that the stressedvowel in managerial is related to the final schwa in manager? American struc-turalist theory had a place for all these generalizations: the morphophonemiccomponent.

Morphophonemics

In Menomini Morphophonemics, Bloomfield presented an analysis that resem-bles a generative derivation: starting from underlying representations, a seriesof rules apply in order to yield phonemic representations. This type of analy-sis, and morphophonemics itself, had a marginal status in structuralist theory.There was very little theorizing done in this area, as opposed to the attention



devoted to phonemic theory. In contrast to the latter, morphophonemics had afreewheeling, anything-goes character, which led, as it turns out, to interest-ing and insightful analyses.10 Why was there such a contrast between the twocomponents? The answer lies in the degree of “reality” attributed to each ofthese levels.

As we have seen, the taxonomic phonemic level was the lowest linguis-tic level recognized in the theory. By assumption, it had to be a level thatcould plausibly be attributed to speakers, including those just learning thelanguage. In keeping with the very restricted conception of psychology andlearning that prevailed at the time, learners were credited with only the mostbasic ability to perform operations of grouping and classification. The variousconstraints placed on the phonemic level were designed to allow a phonemicrepresentation to be easily discovered from the phonetic input available to such alearner.

By contrast, morphophonemics was not given a psychological interpretation.Morphophonemic representations were not necessarily considered to be thingsthat speakers had. According to Anderson (1985: 276), Bloomfield consideredmorphophonemic description to be “an elegant artifact, providing a uniformand concise account of a complex set of facts, but not to be confused withthe actual language capacity of speakers. Only the phonemic forms, and themorphological fact of relations between them, could be considered to have thatstatus.”11

It is significant that Chomsky’s first work in linguistics, his MA thesis(a later version of his BA thesis), is titled Morphophonemics of Modern Hebrew.His later contributions to generative phonology essentially adapted the tech-niques used in morphophonemics – rules and derivations – and placed themat the center of phonological theory (and other components of grammar). Tomake this move, however, Chomsky and Halle had to overcome the argumentsin favor of the taxonomic phoneme. Having shown that the “elegant fiction”of the morphophonemic component was actually real, they now had to showthat the taxonomic phoneme, the rigorous core of phonological theory, was afiction.

Against the taxonomic phoneme

As we saw above, there is no empirical support for a taxonomic phonemiclevel that adheres to the various conditions and restrictions imposed by thepost-Bloomfieldians, suggesting that such a level is unnecessary. In a famousargument, Halle (1959) demonstrated that such a level is also undesirable,because it leads to a loss of generalizations.

Imagine, then, a phonology with three significant levels: a morphophonemic(systematic phonemic) level that was not in dispute; a systematic phonetic level


114 B. Elan Dresher

based on a universal distinctive feature theory; and mediating between them,the taxonomic phonemic level (10).

(10) Three-level phonological component

Underlying forms stored

in lexicon

Morphophonemic level =Systematic phonemic level

Morphophonem

ic

rules


Phonemic forms(Taxonomic) phonemiclevel

Phonetic forms

Phonemic rules

Systematic phonetic level

In Russian, voicing is a contrastive feature that distinguishes pairs of obstru-ent phonemes. Thus, phoneme /t/ is distinct from /d/, /k/ is distinct from /g/, /s/contrasts with /z/, and so on. There is a rule that voices word-final obstruentsif a voiced obstruent follows in the next word. Thus, we find [m’ok l,i] “was(he) getting wet?,” with a k preceding the sonorant l,, but [m’og bi] “were(he) getting wet,” where k voices to g before voiced obstruent b. The voicingrule that changes k to g changes one phoneme to another, and so it must be amorphophonemic rule, applying as in (11a).

(11) Russian voicing applying twice

a. Morphophonemic voicingSystematic phonemic // m’ok bi // // z’ec bi //Voicing m’og bi –Taxonomic phonemic / m’og bi / / z’ec bi /

b. Allophonic voicingTaxonomic phonemic / m’og bi / / z’ec bi /Voicing – z’e�

�bi

Systematic phonetic form [m’og bi] [z’e��

bi]

Three obstruents, / c /, / c /, and / x /, do not have corresponding voiced con-sonants. However, voicing also applies to these segments as well. We have



[z’ec l, i ] “should one burn?,” with voiceless c before the sonorant l,, but [z’e��

bi] “were one to burn,” where ��

is the voiced counterpart to c. Because [��

] isnot a phoneme in its own right, but exists only as an allophone of / c /, this appli-cation of voicing is an allophonic rule, and must be assigned to the componentthat maps phonemic forms into phonetic forms (11b).

Halle argued that the derivation in (11) needlessly splits the voicing ruleinto two (or, alternatively, applies the same rule twice). However, there is noevidence that voicing applies differently in these cases, or that the change occursin two stages rather than just once. Having a taxonomic phonemic level makesit impossible to capture the generalization that there is one voicing rule at workhere, applying equally to all the segments in its purview.12

Without the taxonomic phonemic level, the grammar takes on the form of (6),with only two significant levels of representation: the systematic phonemic, orlexical, level and the systematic phonetic level.

Grammar as a system of knowledge

One of Chomsky’s most fundamental contributions was to reposition linguisticsas a field with implications for the nature of mind and learning. This reorien-tation required a new way of looking at linguistic description, one that wasdiametrically opposed to that prevailing in American linguistics up to thattime.

Bloomfieldian philosophy of science and psychology

Leonard Bloomfield introduced a particularly radical form of behaviorism andscientific empiricism to linguistics. This view included an approach to sciencein general, and to psychology in particular, that together had a great influenceon the development of phonological theory.

With respect to science, Bloomfield and his followers took a view that wasinfluential in the 1920s and 1930s, known as Operationalism, that all sciencemust be framed in terms of statements that describe basic operations, suchas reports of how long it takes an object to travel a certain distance. Puttingthe focus on operations seemed to make the “content” of science out to beobservable matters (measurements and experimental techniques) rather thanobscure-looking “hidden” entities and principles. Theoretical terms and theoryitself, it was thought, could be treated as codes or shorthand for observationsand techniques. Such an approach is incapable of characterizing most importantscientific theories, and the main body of philosophy of science soon abandonedOperationalism.

In American linguistics, however, this general orientation remained influ-ential, and provided a theoretical underpinning to the bottom-up approach tolinguistic analysis. The scientist (here, the linguist) begins with basic data,


116 B. Elan Dresher

that is, a set of utterances in a language (a corpus). These utterances appearto observers (linguists) as a stream of speech. Observers then perform basicoperations to analyze the speech stream: they can segment an utterance intoindividual sounds, or phones; and they can classify the phones into phonemes.These operations can then be repeated at a higher level of analysis, segmentingand classifying strings of phonemes into morphophonemes, then into wordsand phrases, and so on.

American linguistics in fact took this approach even farther than other fields,making it an aim of linguistic theory to devise discovery procedures that wouldautomatically apply the techniques of segmentation and classification to anycorpus and produce an analysis. As Chomsky (1957a) argued, no other sciencehas hoped to arrive at such procedures, which in effect would be an algorithmfor arriving at the correct theory in a particular domain, using only the restricted“data” that operationalist ideology allows.13

These general views about science were complemented by a set of assump-tions about psychology, which amounted to a radical form of behaviorism.Bloomfield and his circle believed that there was no point in attributing a “mind”to any organisms, including humans.14 Rather, behavior is a set of responses tostimuli. Language, then, is simply another form of behavior, verbal behavior, aset of learned responses, or habits.

It followed, on this view, that linguistics had nothing to contribute to psychol-ogy proper. The big questions of psychology – how learning takes place, forexample – were held to be the province of psychologists. The special mandateof linguistics was to study verbal behavior proper – to describe (and merelythat) the way utterances are put together and used (primarily, it was assumed)in communication. Thus, linguists in this framework were given to say that lin-guistics should be free of psychology, and should make no assumptions aboutpsychology. This statement, given the above, was quite obviously disingenuous.What was meant, however, is that linguists need not get involved in psycho-logical speculation, beyond the background assumptions sketched above. Thatis, internal to a linguistic description there is no need to invoke psychology.Nevertheless, the analysis is couched in a framework that is heavily indebtedto a specific model of psychology.

The combination of scientific and psychological assumptions that formed thebackground of neo-Bloomfieldian linguistics placed narrow limits on the typesof linguistic theory that could be entertained. Moreover, they tended to draw thefocus of inquiry away from evaluating how successful particular theories werein accounting for the facts of language – capturing significant generalizations,accounting for how language is acquired, and so on – and emphasized insteadconformity to what were in effect a priori and unmotivated restrictions on theform of a linguistic theory, whatever the consequences might be for particularanalyses.



Chomsky’s philosophy of science and psychology

Chomsky’s approach represented a radical break with Bloomfieldian thoughtboth with respect to science and psychology. With respect to science, Chomskybegan with the premise that a linguistic description is a hypothesis about howthe data is organized, and thus is a scientific theory of the grammar. He pointedout that no other science has developed, or even seeks to develop, discoveryprocedures – a set of automatic procedures applied to basic data that results in atheory of the data. Rather, scientists arrive at hypotheses about nature howeverthey can. What can be expected of a scientific theory is not an account, let alonejustification, of how it was arrived at, but criteria for assessing how good a theoryit is. That is, rather than discovery procedures, linguists should be concernedwith developing evaluation procedures, means by which to compare competingtheories (grammatical descriptions) with a view to determining which is thebetter theory.

With respect to psychology, Chomsky argued that behaviorism chose thewrong target. There is not, and likely will not be, a theory of human behavior.15

But it may be possible to characterize aspects of cognition, or knowledge,that contribute to behavior. He therefore proposed that the proper object oflinguistics is not verbal behavior, which is the product of diverse systems, butrather knowledge of language. Chomsky proposed that a grammar is actually atheory of the knowledge that native speakers have of their language.

Learnability and Universal Grammar

If a grammar is an account of knowledge, the question immediately arises asto how native speakers come to have this knowledge. As he has argued atlength (notably in Chomsky 1975), one cannot consider learning in a domainapart from the cognitive principles that the learner brings to it. In the case oflanguage, where many considerations point to a specialized ability shared by allhumans, there is no general learning theory we can appeal to. Rather, linguisticinquiry itself must determine, first, what is actually acquired, and second, whatcognitive principles learners employ. The answer to the first question will bea grammar of the language that has been acquired. The answer to the secondquestion is a set of universal principles collectively called Universal Grammar(UG). Whether a particular grammar is easy to learn or difficult depends on thecombination of accessibility to relevant data and the nature of UG.

SPE begins with the sentence, “The goal of the descriptive study of a languageis the construction of a grammar.” Much of SPE is devoted to constructing thephonological part of the grammar of English. Freed from restrictions on the rela-tion between phonemes and their surface allophones and from the requirementthat morphological relations may not enter into a phonological analysis, the


118 B. Elan Dresher

phonology of English that emerges in SPE is radically different from previousaccounts.

As the opening sentence indicates, SPE is much more concerned with theparticular grammar of English than with how this grammar could be acquired.Nevertheless, the phonology of English presented in SPE is set within a generaltheory of phonology which is intended to provide a first attempt at a theory ofUG in the domain of phonology.

The main mechanism proposed by SPE for guiding a learner to the correctgrammar is the evaluation measure. The evaluation measure, assumed to bepart of UG, assigns a higher value to a rule that uses fewer features over arule that uses more features. This measure reflects the fact that features capturenatural classes: therefore, the fewer features, the more general the class. A rulethat applies to a more general class is generally preferred to one that delimits anarrower class. That is, a simpler rule is preferred over a more complex one.

While such an evaluation measure can guide a learner in making local deci-sions between similar rules that differ in their formal simplicity, it is of littlehelp in allowing a learner to choose between sets of rules, or even entire com-peting grammars. With respect to this problem, there are parallels in the earlydevelopment of phonological and syntactic theory. In both cases, the theoriesinherited from structuralism were too descriptively limited: they did not possessthe resources to provide a descriptively adequate account of their subject matter.In both cases, therefore, the emphasis at first was on expanding the descriptiveresources of linguistic theory. In the case of syntax, this meant exploring thepower of transformations; in the case of phonology, the power of derivationsand the rule formalism introduced by SPE.

In syntax, a concern with constraining the theory arose in the late 1960sand eventually came to dominate thinking in the field, at least in the partof the field associated with Chomsky. By this time, however, Chomsky hadlargely given up working on phonology, and was not involved with subsequentdevelopments.16 Many of these developments took as their starting point thetheory of SPE. Here I will only highlight certain issues that particularly pertain tolearnability.17

Abstractness of phonology

The underlying forms posited by SPE are in general rather abstract with respectto their phonetic surface forms. Much ink has been shed over the extent towhich such abstractness is learnable, or whether there should be constraints onabstractness.18 It is an empirical question whether phonological systems obeyconstraints limiting the “distance” between underlying and surface forms. Ifthey do, we expect to find evidence for this from the patterning of the phonology



itself. However, in some cases proposals to constrain abstractness have tendedto be based not on empirical evidence, but, rather, on a priori assumptionsconcerning what is learnable.

There is no particular reason to suppose that there is a simple relationbetween abstractness and learnability. Consider again, for example, the deriva-tion in (4). The rule of Flapping contributes to create a considerable mis-match between the underlying and surface forms, as we have seen. At thesame time, Flapping obscures the contexts of Raising and Lengthening, bymaking it appear that Raising applies before a voiced consonant (the flap [ɾ])in writer and that Lengthening fails before a voiced consonant in the sameform. This derivation is ruled out by theories that place restrictions on abstract-ness. Kiparsky (1973), while not ruling out such derivations, proposed thatrules that are opaque by his definition (i.e. contradicted by surface forms)are difficult to learn. Intuitively, one might expect this to be so: a learnerwho has not yet acquired the rule that raises /aj / to [�j] before voicelessconsonants might be misled by hearing forms like writer ([r�jɾər]), wherethe raising has taken place, though the following consonant is voiced on thesurface.

Nevertheless, it is not obvious that this grammar is difficult to learn. The rulesof Raising and Centralization apply without complicating factors in the wordswrite and ride, as well as many other words of English. Similarly, the rule ofFlapping applies transparently in many words, such as sit ∼ sitter and dad ∼daddy. A learner who has learned this much of the grammar already has themain ingredients of the derivation except for the relative ordering of Flappingand the other rules, which must be acquired on the basis of forms like writerand rider.19

Just as it is not the case that abstractness necessarily leads to learnabilityproblems, neither is it the case that reducing abstractness necessarily makeslearning easier. For example, in Seri (Marlett 1981) there is a set of vowel-initial words that behave as if they begin with a consonant. Marlett proposesthat such words have an underlying initial consonant, designated C, which isdeleted toward the end of the derivation. This is an abstract analysis, because itposits a segment that is present in underlying and intermediate representationsthat is never audible at the surface.

Marlett and Stemberger (1983) present a different analysis. They adopt anonlinear theory of phonology in which phonological representations consistof different tiers. One of the tiers is a CV-tier, or “skeleton,” which relatessegmental features to syllable structure. Rather than posit an abstract underlyingconsonant, they propose that the exceptional verbs have an initial C position onthe CV-tier that has no counterpart on the feature tiers. The two analyses arepresented schematically in (12).


120 B. Elan Dresher

(12) Two analyses of Seri exceptional verbs

a. Abstract CPhonemic / CaX /RulesC-deletion aXPhonetic [aX]

b. Empty CCV-tier C V C

| |Feature tiers a X

Marlett and Stemberger comment that the Empty C analysis is superior to theAbstract C analysis because it is less abstract, since the empty C is present at thesurface. However, if we focus on learnability and ask what would lead a learnerto posit the empty C in (12b), we find that it is exactly the same evidence thatwould lead the learner to an abstract C in (12a). Since abstractness is a relationbetween levels of a derivation, it is no surprise that one can reduce abstractnessby enriching representations. In (12), the “vertical” relation between a phonemicC and its null phonetic realization is replaced by a “horizontal” relation betweena C on one tier and a null representation on the feature tier(s). The analysis in(12b) raises interesting issues about the nature of representations, but does littleto advance the problem of learnability.20

Markedness and other substantive principles

While abstractness in itself is not necessarily a special problem for learnability,when combined with the otherwise rather loosely constrained apparatus of SPEa real problem for learnability does arise. In short, the theory of SPE does notprovide an adequate answer as to how learners are able to arrive at the actualgrammar of their language and not any number of other grammars.

One shortcoming of the SPE theory, recognized within SPE itself, is thatthe theory is overly formal: the evaluation measure counts only the number ofsymbols in a rule, without taking account of intrinsic properties of features orof phonological processes. Thus, a three-vowel system with vowels / i a u / isvery common, but one with the vowels / u ε æ / is unheard of. Similarly, a rulechanging i to u is much more common than one changing i to i , even thoughu differs from i by two features ([back] and [round]) and i differs from i byonly one feature ([back]). To remedy this fundamental theoretical inadequacyChomsky and Halle introduce a theory of markedness. Taking up and extendingideas from the Prague School, they propose that certain feature values andcombinations are unmarked, or default, whereas others are marked, and entail



a greater cost to the grammar. By making phonological rules and inventoriessensitive to markedness considerations, one can explain why certain segmentsare more common than others, and why certain rules are more highly valued thanother rules with the same number of symbols. Markedness theory contributesto learnability by giving further structure to the hypothesis space of possiblegrammars.21

Asymmetries in intrinsic content also arise in the relationship between phono-logical changes and their contexts. For example, rules assimilating one or morefeatures of a segment to features in the environment are very common. Anextension of markedness theory would also take account of the naturalness ofrules.

I have suggested that various proposals to limit abstractness have notimproved the learnability of grammar. On the contrary, it can be argued thatabstractness of the right sort actually improves learnability, by working togetherwith markedness and naturalness to limit the choices available to a learner.

Dresher (1981b) adduces an argument along these lines from Old English.The verb eotan “to eat” can be shown to have the underlying stem vowel /e/. Thisvowel is raised to [i] in the present indicative second and third person singularforms ites and iteθ. Dresher proposes that the suffixes in these forms derive fromunderlying /+is/ and /+iθ/, respectively, and the forms in question are derivedby means of two rules, given informally in (13) and (14). The derivations areshown in (15).

(13) e-RaisingStressed e raises to i when an i follows in the next syllable.

(14) i-LoweringUnstressed i lowers to [e] when it follows a light syllable (a syllablecontaining a short vowel and followed by a single consonant).

(15) Derivation of forms with stem vowel i

Underlying / et+is / / et+iθ /e-Raising itis itiθi-Lowering ites iteθPhonetic [ites] [iteθ]

This analysis is supported by a web of other evidence (see Dresher 1985 fordetails). Assuming here the correctness of the analysis, the question we wishto answer is how a learner of Old English could arrive at it. In particular, whatwould lead a learner to suppose that the suffixes are / +is / and / +iθ / when theyalways appear as [+es] and [+eθ]?

Learners would know that unstressed e does not in general cause raising ofa preceding i: there are many words like the present subjunctive eten “we / you


122 B. Elan Dresher

(pl.) / they eat.” It is possible that raising in the present singular is simply anexception, conditioned by morphology, just as the English vowel change fromfoot to feet must be attributed to an irregular rule that operates in the plural.While such an analysis cannot be excluded, the assumption of SPE is that it isnot preferred by the learning theory. In the formulation of Postal (1968), theSPE theory incorporates the Naturalness Condition.

(16) The Naturalness Condition (Postal 1968)Phonological classifications are preferred to morphological or arbi-trary classifications at all levels of the phonology.

The Naturalness Condition instructs the learner to seek a phonological solutionbefore falling back on a morphological one.

In a learning theory which values simple and natural rules learners couldproceed from the change to make hypotheses about the context. The raisingof / e / to [i] involves a change of one feature, [–high], to [+high]. The mostfavored context for such a change is in the vicinity of another [+high] feature. Inour case, there is no such segment visible on the surface; however, the vowel -ein the suffixes -es and -eθ is only one feature away from a high vowel, namely -i.Thus, even lacking other evidence, learners equipped with a learning theory ofthe kind sketched above may suspect at an early stage that the suffix vowel in-es and -eθ may be a disguised / i /. In this case, the suspicion will be supportedby evidence from other parts of the phonology.

Notice that this solution would not be possible without abstractness. Byallowing a certain distance between phonemic and phonetic forms, we can takeadvantage of rule ordering to formulate rules that are maximally simple andnatural. In this way, abstractness contributes to learnability.


6 Universal aspects of word learning

Lila Gleitman and Cynthia Fisher

Suppose instead of saying Dutch we had said Clashes with the wallpaper; Ithought you liked abstract work, Never saw it before, Tilted, Hanging too low,Beautiful, Hideous, Remember our camping trip last summer?

(Chomsky 1959; Review of B. F. Skinner’s Verbal Learning)

Most of the action in linguistic theory under Chomsky’s aegis has focused onquestions of how words are put together into sentences rather than on the wordsthemselves. Fair enough: the universal hallmark of human language is its para-metrically organized combinatorial structure. In linguistic systems, the individ-ual word classes (for instance, the nouns, the verbs, the complementizers) playparticular and crucial structural roles; rules and parameters directly implicatethese classes. But what about the items that comprise these classes? It has beenwidely maintained that these individual atoms or particles of language play nocentral role in its core, or, to use recent terminology, in “language narrowlyconceived” (Hauser, Chomsky & Fitch 2002). No distributional property ofEnglish worth its salt is dependent on whether some particular noun – let ussay elephant rather than rhinoceros or gnu – appears in it.1 Rather, what wordsseem to be relevant to is the conceptual system that interfaces with language.In this latter domain, relevant to “language broadly conceived,” the differencebetween an elephant and a gnu really matters.

Similarly, the linguistic-theoretical study of language acquisition has focusedprimarily on the nature and setting of syntactic parameters rather than on theevolving character of the lexicon (e.g. Manzini & Wexler 1987; Lightfoot 1999;Baker 2001). Again, this is no surprise. Acquisition of the combinatorial fea-tures of a language poses a classical poverty-of-the-stimulus problem, requiringthe learner to extract abstract organizing principles from input structured only(or almost only) as morpheme sequences (see, particularly, Chomsky 1986).In contrast, a first intuition is that acquiring words and their meanings canbe fully accounted for by a procedure that associates the forms (say, the sound“elephant”) with their meanings (here, the concept elephant) in consequenceof observing the referential contingencies for the word’s use (say, visible pres-ence of an elephant). Here is this claim, as put by John Locke (1690):

123


124 Lila Gleitman and Cynthia Fisher

(1) If we will observe how children learn languages, we shall find that . . . peopleordinarily show them the thing whereof they would have them have the idea, and thenrepeat to them the name that stands for it, as “white,” “sweet,” “milk,” “sugar,” “cat,”“dog,” (Book 3, IX, 9)

In this chapter we will revisit lexical learning in aspects that seem to be rele-vant to Chomsky’s program for understanding language and its users. We willfocus on four broad issues. First, we show that the method of word-to-worldpairing advocated by Locke in (1) is too weak, taken alone, to account for therobustness of word learning. Second, we discuss further data sources – linguis-tic data sources – that in principle can enrich and constrain this acquisitionprocedure. Third, we describe what is known of the circumstances in whichchildren recruit and exploit these various data sources. Fourth and finally, weallude to the accumulating evidence that the word-learning procedure leaves itsfootprint in the mature mental representation of language: grammars are heavilylexicalized, in part because the learning procedure for words has – necessarily –built complex (syntactically organized) lexical structures.

Before beginning, we want to make clear the sense of “word learning” that wewill discuss. We restrict ourselves to the mapping problem: how one comes toknow, if exposed to French, that chien is the phonological form that expresses theconcept dog while voir expresses to see ; whereas if exposed to Italian, that itis cane and vedere. We leave aside altogether the question of where the conceptsthemselves come from, stipulating only that, to a useful approximation, theseconcepts are in place to support the word-learning procedure. That is, acquiringthe meaning for dog and see requires that the learner antecedently be able toentertain these concepts2 (Fodor 1983). Our question – the mapping problemfor vocabulary acquisition – is how the child decides which sound goes withwhich meaning.

The robustness of learning to input variation

Locke’s dictum (1) is that the correlation between word use and the specificsof the reference world is both necessary and sufficient to account for wordlearning. It follows from this that words could not be acquired (the mappingproblem could not be solved) except under conditions in which words wereuttered in the presence of their referents. We ask in this section whether thisprecondition is met in children’s learning environment; and, if it is, whetherthis precondition is sufficient as well as necessary.

The contingencies for a word’s use

Perhaps the central difficulty with Locke’s approach was noted by Chomsky(1959), who challenged empiricist speculation on just this point; hence the


Universal aspects of word learning 125

epigraph to the present chapter. The trouble is that one does not always or evenparticularly often say “Dutch” on viewing a Rembrandt painting. Conversa-tion, even from mothers to babies, is not a running commentary on the objects,events, properties, and relations presently on exhibit in the world. No mothercarefully utters “open” every time she opens the door; worse, “open” is fre-quently uttered – even systematically so – when the door is shut (“Help! I’mlocked in this bathroom! Please someone come and open the door!”). In the nextmain section of this chapter we will document this “stimulus-free” property oflanguage use. The problem is not that utterances have no relevance to matters athand. It is rather that the concept relevance is so broad that it places very littleconstraint on what is likely to be said, given some circumstance. We shouldnote that the issues here look even more ominous from the perspective of Quine(1960), who famously noticed that indefinitely many construals can be put onany observation of the world (say, a rabbit-observation). Chomsky’s objectionsseem more to the point, as he grants constraints on interpretation deriving fromhuman conceptual structure and from conversational relevance; and yet still theword-to-world contingencies appear to be so flexible as to render learning byobservation intractable.

Cross-situational observation

The standard response to the problem just discussed is that the child doesn’thave to learn from a single observation. Rather, over successive observations,probabilistic relations between word and world will converge to support wordlearning. After all, in the end elephant is uttered more in elephant situationsthan in zebra situations (Pinker 1984). One potential pitfall for a learning devicereliant on these correlations is the ornate differentiating series of observationsthat would be required if it is not to go hopelessly wrong, a difficulty that Lockeseems to acknowledge by his very choice of examples: in (1) whiteness is anattribute of both the sugar and the milk, and sweetness of both the sugar and(if one is lucky) one’s mother. An even more difficult problem is that such aprocedure does not seem to comport well with the actual learning facts. Forone thing, a statistical learning procedure that maps between word sounds andevents must be errorful over its course, involving significant backtracking andrevision, and so predicts some proportion of howlers along the way. Childrenshould sometimes mix up the milk and the sugar with their mothers, at least inspeech acts. Moreover, vocabulary acquisition should overall be a slow process,since an accumulation of observations is required to warrant conclusions abouteach item (and the more variable the word-to-world circumstances, the slowershould be the learning). Yet learning proceeds at rates up to ten new itemsper day, almost completely without the predicted howlers. Children seem tobe drawing the right conclusions about the meanings and referents of words



based on one or a very few exposures. In Carey’s (1978) terms, they are “fastmappers.”

Populations deprived of information; colorless green ideas

Just as no child receives all the sentences of English as the condition for learningits grammar, so no child observes all the potential referents as a condition foracquiring word meanings. Totally disjoint dog observations underlie individu-als’ acquisition of the word dog. You see Fido and Spot, I see Rex and Ginger,yet for each of us a generalization results such that we consensually partitionthe whole world into the dogs and the non-dogs. The British Empiricists wereamong the first to ask just how far observational environments could diverge,to support the learning of the same word. They suggested that relevant test bedsfor the learning by observation hypothesis can come from populations who aresystematically deprived of certain opportunities to observe the world. Here isthis suggestion, as offered by David Hume:

(2) . . . wherever by any accident the faculties which give rise to any impression areobstructed in their operations, as when one is born blind or deaf, not only the impressionsare lost, but also their correspondent ideas; so that there never appear in the mind theleast trace of either of them. (1739/1978: 49).

For example, the congenitally blind do not observe redness or seeing. Thereforeby hypothesis they could not acquire the concepts red and see . It follows, ifword learning requires a mapping between instantiations of a concept and thehearing of a word (qua phonological object), then a blind child would have nobasis for learning the meaning of words that express vision-related concepts.Two cases studied by Landau and Gleitman (1985) were vision verbs (look,see) and color nouns and adjectives (color, green, red). Sighted blindfoldedthree-year-olds told to “Look up!” turned their faces skyward, suggesting thatthey interpreted look to implicate vision in particular. This interpretation isn’ttrue of the blind: a blind three-year-old given the same command raises herhands skyward instead of her face, suggesting that for her the term is connectedto the manual sense.

So far so good for Locke and Hume: the difference in observational opportu-nities leads the blind and sighted to different interpretations of the same term.Successful communication from mother to blind child using this term oftenoccurred just when the objects to be “looked at” were in the learner’s hands,licensing a physical contact interpretation of blind looking. However, severalcommon verbs used by the mother to the blind child shared the property ofbeing uttered – and even more systematically than look – when the child had arelevant object in hand, including hold, give, put, and play.

Moreover, the blind child’s interpretation of look goes beyond manual con-tact. An informative manipulation was to say “You can touch that table but



don’t look at it.” If look means touch to the blind, this command is incoherentand therefore cannot be obeyed; but instead the blind child gingerly taps orscratches at the table in response to this command. Subsequently told “Nowyou can look at it,” the child explores the surfaces of the table manually. Basedon this kind of evidence, Landau and Gleitman (1985) concluded that blindlook (a) semantically differs from sighted look by implicating a different sensemodality, but (b) semantically resembles sighted look, and differs from hold,touch, etc., in being a term of perception. To be sure, in order to look a blindperson must touch; but that does not imply that look means – or “just means” –touch. (And predictably, “You can look at this table but don’t touch it” elicitsconfused complaints from the blind child.)

Summarizing, we can easily account for the blind child’s failure to map lookonto the visual modality from an orthodox associative perspective on wordlearning that focuses on the necessity of extralinguistic observation. But thisperspective cannot so easily explain how blind – and sighted – learners hit uponlooking as a perceptual rather than as a contact term. Again, these findingssuggest that word-to-world pairing is insufficient as a full explanation of wordlearning. Both populations know too much, from too little information aboutthe world. Chomsky (1986) termed this “Plato’s Problem.”

The blind child’s understanding of color terms offers a similar insight: Landauand Gleitman’s (1985) blind preschool-aged informant knew that (a) color is thesupernym for a subset of adjectival terms including green and red, but not cleanand happy; and (b) the color terms apply only to concrete objects. Asked “Can adog be blue?” the blind child at five years of age responded with different colorterms: “A dog is not even blue. It’s gold or brown or something else.” Asked“Can an idea be green?” the child responded “Really isn’t green; really justtalked about – no color but we think about it in our mind.” That is, blue may notbe an actual attribute of dogs; but green is a category error as applied to ideas.These findings display the remarkable resilience of semantic acquisition overvariations of input: lacking the ordinarily relevant observations that supportsolution of the mapping problem for visual terms, the blind are not helpless todo the same. But then what is the foundation for this learning?

Hard words

Many of the words that mothers frequently utter to their infants are so divorcedfrom straightforward perception that it is hard to see how observation couldpossibly be available to support their acquisition. Even supposing (somewhatcontroversially) that the learner by the age of two or three is capable of under-standing the concept we express as think, it is quite difficult to imagine circum-stances, short of visits to the Rodin Museum, that would bring it to mind aswhat a conversation is about. If we point to some group of people who truly are



thinking, and contrast them with another set of people who truly are not think-ing, what good can it do the learner to gaze upon these contrasting people-sets?Their thinking is happening inside their heads, invisible to any observer. Itemsof this sort have posed important challenges to the view that the mapping prob-lem can be accomplished by a machinery responsive solely to the observablecontingencies for a word’s use. “Show them the thing . . .” seems a very muchless compelling method for this kind of item than it does for the elephants.

More generally, the trouble with verbs and other predicate terms is that they’reabstract, and therefore much less obviously displayed in the flow of events.Verb meanings depend not only on the events in view, but also on a choice ofperspectives on events (Clark 1990; Gleitman 1990; Gleitman, Gleitman, Miller& Ostrin 1996; Pinker 1989, inter alia). At the extreme, perspective-changingverbs like chase and flee pose a problem of principle for any theory of lexicalacquisition that relies solely on word-to-world mapping. Verbs within thesepairs describe the same events, differing only in their focus on the perspective ofone or the other participant in that event. This focus difference is unobservable,residing only in the speaker’s head. As further examples, one cannot say of ascene that one of its participants is giving if nobody gets, and every time oneputs the cup on the table, the cup also lands on the table (Gleitman 1990).

Inconvenient facts about word learning

Children do not learn every word they hear. Input frequency does not even beginto explain this fact. Perhaps we need no complex theory to explain why no child’sfirst word is the despite its frequency in maternal speech. But a timing differencefor object terms versus action terms (and its surface correlate, nouns versusverbs) requires a little further theorizing. Many studies show that children’s earlyproduction vocabulary is dominated by concrete nouns – names for objects andpeople in particular (see for reviews Gentner & Boroditsky 2001; Woodward &Markman 1998). This is true in languages other than English, even in languageslike Italian that possess surface properties conducive to verb learning, includingthe omission of inferable noun phrases. The same bias toward learning objectnames is present in the earliest language comprehension as well. Novel wordspresented in object-manipulation contexts cause one-year-olds to focus on thekinds of similarity across objects that can indicate shared category membership(Waxman & Markow 1995). When a new word is presented, the object-kindinterpretation is often so salient that it’s difficult to get children to arrive at anyother interpretation (Bloom 2000; Gentner 1982).

At first glance, this bias does not seem to be in conflict with the observationaltheory of word learning. It could be – this surely is true in the limit – that someconcepts are more difficult to entertain than others, and therefore are simplyunavailable to the infant mind. The relational notions that verbs typically express



might be harder to grasp, or less salient to infants, than are notions of object kind.If the relevant world observations are potentially available but uninterpretable,then the words can’t be learned. This account of the late appearance of verbsrelative to nouns would thus accord perfectly well with the Lockian prediction.However, there is a different account of this striking, universal, input–outputdisparity for the efficient acquisition of nouns and verbs: the acquisition ofnouns and verbs may require different kinds of information, and the informationsources themselves may become available at different developmental moments.We turn now to evidence for just such a position.

Linguistic and conceptual supports for vocabulary acquisition

How can the robustness of word learning be understood in the face of thevagaries of ordinary experience and the vast array of reasonable interpretationsof what has been said? We will argue that vocabulary acquisition is not of apiece. Some words are necessarily learned before others. The initial learning ofconcrete nouns sets in motion a process that makes possible the efficient learningof less concrete words; bits of the lexicon and grammar of the exposure languageare acquired in a succession of causally interlocking steps. Learners constructthe linguistic ladder, so to speak, while they are climbing it.

Word-to-world mapping: showing them “the thing” suffices for somewords, but not for think – or thing

Some words are less obscure than others in the flow of experience made avail-able for our inspection. Evidence for systematic variations in the recoverabilityof various words’ meanings from world context alone comes from studies byGillette, Gleitman, Gleitman, and Lederer (1999). Their interest was in under-standing how information structure in the input influences solution of the map-ping problem, apart from whatever role developmental differences in mentalitymay play. Therefore they used adults to simulate vocabulary learning undervarious informational circumstances.

The first step in these investigations was to understand the limits of word-to-world pairing: solving the mapping problem by using observed scenes as thesole clue. To do so, the investigators showed their adult subjects brief videoclips(about 45 seconds in length) of mothers and toddlers playing with toys andconversing. The soundtracks were removed from these video clips, and a “beep”was inserted in each clip at the moment when the mother had uttered a particulartarget word. The targets were the 24 nouns (e.g. ball, hand, hat) and 24 verbs(e.g. push, come, look) most frequently produced by mothers in the corporafrom which these clips were drawn. For each word, the adult observers weretold they would see six videotaped occasions of the same word’s use in a row.



Thus they had some opportunity for cross-situational observation. Their taskwas to identify each “mystery word” using this accumulating evidence. Withonly these scene observations as evidence, adults correctly identified three timesas many of the mothers’ nouns (about 45 percent) as of their verbs (about15 percent).

Success rates in this task could be predicted by other adults’ judgments ofthe imageability (concreteness) of each word. On average, the common nounsin the mothers’ speech were judged more imageable than the common verbs,and variability in judged imageability was a better predictor than the noun/verbdistinction of which words were successfully induced from observation of thescenes. The most concrete of the target verbs (e.g. throw) were identified morefrequently than the most abstract. Those judged most abstract, including thinkand know, were never guessed correctly by any subject. Subsequent studieshave begun to refine the notion of concreteness that determines which wordsare relatively easy to learn from observation alone. For example, Kako andGleitman (in prep.) found that words for basic-level categories of whole objects(e.g. elephant) are strikingly easier to identify based on observations of theircircumstances of use alone than are abstract nouns (e.g. thing) or part terms(e.g. the elephant’s trunk).

These findings (see also Snedeker & Gleitman in press) yield a simple expla-nation for the probabilistic noun advantage in infants’ first vocabularies. Theadult subjects in Gillette et al.’s (1999) studies had already grasped the conceptslexicalized by all the English words to be guessed in the study. Nevertheless,only the most concrete words were successfully identified from observing theextralinguistic contexts alone. The most concrete words, including a usefulvocabulary of names for things, are just those for which linguistically unaidedobservation is likely to be informative.

These data confirm that the solution to the mapping problem may start withLocke’s procedure but cannot end there. Observation of the thing is sufficientfor the acquisition of some (e.g. elephant) but not all of our words (e.g. thing,trunk, think). The true beginner can only try to observe elements in the worldthat systematically covary with the use of particular words. This leads to successin those cases in which the word’s meaning is concrete enough to be readilyobservable in the flow of events: mostly nouns, but also a heterogeneous set ofother words.

Sentence-to-world mapping

How does the child move beyond an initial concrete, largely nominal,vocabulary? To learn less concrete (less observable) terms, the learner needsother kinds of evidence – linguistic evidence, bootstrapped from (grounded by)the previously acquired vocabulary of concrete words.3



The view known as syntactic bootstrapping proposes that the interpretationof verbs and other predicate terms is guided by information about the structureof the sentence in which the verb appears (Landau & Gleitman 1985; Gleitman1990; Fisher 1996). Most generally, this view proposes that word learning afterthe first steps proceeds by sentence-to-world pairing rather than merely byword-to-world pairing.

To illustrate, let us return to the Gillette et al. (1999) “human simulations”described earlier. These investigators repeated their experiment, asking adultsto identify verbs spoken to young children based on various combinations oflinguistic and extralinguistic information. Adults were much more accurate inguessing which verb the mother said to her child when given information aboutthe sentences in which the verb had occurred. When given a list of the nounsthat occurred in the sentence (alphabetized to remove word-order information),along with the scene in which the verb was produced, subjects’ guesses weresignificantly more accurate than when given the scene alone. Subjects also prof-ited from more explicit syntactic information about the verbs’ original contextsof use, even when denied observation of the scene: presented only with a set ofsentences in which all the content words were replaced with nonsense words(e.g. Can ver gorp litch on the fulgar?; much as in Carroll’s poem Jabber-wocky), subjects were significantly more accurate in guessing the verbs thanwhen they saw the scenes, or even when they saw the scenes plus an alpha-betized list of cooccurring nouns. When presented with the complete sentencecontexts, with only the verb replaced by a nonsense word (e.g. Can you gorpMarkie on the phone?), subjects’ guesses were quite accurate even withoutaccess to the scenes, and nearly perfect with both sentence and scene.

Why would syntactic information so strongly guide semantic inferences?Verbs vary in their syntactic privileges (i.e. the number, type, and positioningof their associated phrases). These variations are systematically related to theverbs’ meanings (Chomsky 1981a; Fisher, Gleitman & Gleitman 1991; Gleit-man 1990; Grimshaw 1990; Jackendoff 1983; Rappaport, Hovav & Levin 1988;Pinker 1989, inter alia). A verb that describes the motion of an object will usu-ally occur with a noun phrase that specifies that object; a verb that describes anaction on an object will typically accept two noun phrases (i.e. be transitive); averb that describes the transfer of an object from one position to another willtake three arguments. Similarly sensible patterns appear for argument type: seecan take a noun phrase as its complement because we can see objects, but alsocan take a sentence complement because we can perceive states of affairs.

Such syntactic–semantic correspondence patterns show striking regularitiesacross languages (Baker 2001; Croft 1990; Dowty 1991). These crosslinguisticregularities have long been taken to be primary data for linguistic theoriesto explain, leading to principles such as the theta criterion and the projectionprinciple (Chomsky 1981a), which jointly state that the nouns in sentences must



be licensed by the right kind of predicate (one that can assign them a thematicor “theta” role), and that clause structure must be projected from lexical entries.Similarly, unlearned constraints linking thematic roles such as agent and themeto grammatical functions like subject and object have been proposed to explaincrosslinguistic regularities in the assignments of semantic roles to sentencepositions. Causal agents, for example, overwhelmingly appear as grammaticalsubjects across languages (Baker 2001; Keenan 1976).

Based on these systematic links between syntax and meaning, the adults inGillette et al.’s (1999) studies, or a suitably constructed young learner, canconsult each verb’s sentence structure to glean information about its meaning.The observed sentence structure, by specifying how many and what types ofarguments are being selected by the verb, provides a kind of “linguistic zoomlens” to help the learner detect what is currently being expressed about anongoing event or a state or relation. The set of such structures associated witha verb, across usages, is a complex function of its full expressive range (wediscuss these issues further on p. 138, below).

Linguistic evidence aids identification of abstract nouns as well. Kako andGleitman (in prep.) found that the inductive advantage for basic-level objectkinds was reduced when linguistic information was added to or substituted forthe scene information (Sorf the RENCK’s reb? or See the RENCK’s trunk?).Consistent with this finding, several prior studies have found that abstract,superordinate, or part nouns are typically introduced into the conversation ininformative linguistic contexts (e.g. This is the bear, here are his ears; Here’sa dog, a cat, and a horse, they’re all animals; see Shipley, Kuhn & Madden1983; Callanan 1985).

Inferences from syntax to meaning will presumably differ in their mechanicsfor abstract nouns and for verbs. To a considerable degree, however, sentencestructures will be informative insofar as they convey information about thepredicate-argument structure of their meanings, for argument-taking nominalsof various sorts (e.g. John’s shoe; the fact that Bill likes ham), nominal argu-ments of known verbs (e.g. feeding the ferret), unknown verbs (e.g. she adoresham), or other argument-taking predicates (e.g. the cat is on the mat).

Summary: the information base for word learning

The “Human simulation” studies just discussed tell us about the informationstructure of the input: more than one kind of information is available for vocab-ulary learning, and these information sources are more or less informativedepending on the kind of word being learned. For the case of basic-level namesfor things, reference is fairly easy to determine from unaided inspection of thescene, whereas there is almost no information to be gained from these words’licensed positions in sentences (other than the fact that they are nouns). It is easy



to see why. Tens of thousands of English nouns appear in the linguistic con-text “the gorp,” so this context hardly narrows the search-space for mapping;in contrast, it is relatively easy to observe, say, a horse or a flower or a forkin the situational context. The meanings of more abstract words, includingmost verbs, are harder to identify in the flow of events, but have more infor-mative linguistic contexts. At the far end of this abstractness continuum arethe credal verbs, such as think and know, for which the situational observa-tion is of almost no value, while the linguistic-syntactic information is hugelyinformative for these very cases (Gillette et al. 1999; Snedeker & Gleitman inpress).

Children’s use of multiple cues in word learning

Important as it is to determine the potential informativeness of the multiple cuesto word meaning available in the input, it still remains to demonstrate how andwhen learners are responsive to them.

The meanings to be communicated, and their systematic mappingonto linguistic expressions, arise independently of exposure toany language

In advance of language learning, infants during the first year of life naturallyfactor their representations of events into conceptual predicates and arguments(Bloom 2000; Fisher & Gleitman 2002). Some of the most striking evidencethat the structure of human cognition yields a language-appropriate divisionof our thoughts into predicates and arguments comes from learners who areisolated from ordinary exposure to a language and therefore have to invent oneon their own.

Most deaf children are born to hearing parents who do not sign, and thereforethe children may not come into contact with gestural languages for years (New-port 1990). Deaf children with no available language model spontaneouslyinvent gesture systems called “Home Sign” (Feldman, Goldin-Meadow &Gleitman 1978; Goldin-Meadow 2003). Remarkably, though these children areisolated from exposure to any conventional language, their home sign systemspartition their experience into the same pieces that characterize the elements ofsentences in Italian, Inuktitut, and English. Specifically, home sign systems havenouns and verbs, distinguishable from each other by their positions in the chil-dren’s gesture sequences and by their distinctive iconic properties. Moreover,and especially pertinent to the issues that we have been discussing, sentence-like combinations of these gestures vary in both the number and positioningof the nouns as a function of what their verbs mean. Systematically appearingwith each verb in a child’s home sign system are other signs spelling out the



thematic roles required by the logic of the verb: the agent of the act, the patientor thing affected, and so forth.

The nature of this relationship is easy to see from a few examples: Because“crying” involves only a single participant (the crier), a verb with this meaningappears with only one nominal argument. Because “tapping” has two partici-pants, the tapper and the thing tapped, such verbs appear with two nominal argu-ments. Because “giving” requires a giver, a getter, and a gift, this verb showsup with three nominal phrases. As mentioned earlier, these semantic functionsof the nouns vis a vis the verbs are known as their thematic or semantic or thetaroles (Chomsky 1981a). The same fundamental relationships between verbmeaning and nominal arguments surface in much the same way in the speechof children who are acquiring a conventional language, and in the gestures oflinguistically isolated children who must invent one for themselves.4

In addition, the nouns occurring with each verb do not appear haphazardlyto either side of the verb. The isolated deaf children adopt systematic gestureorders, such as routinely signing undergoers immediately before verbs, (transi-tive) agents following verbs, and intransitive actors before verbs. Thus, a homesigner who produced “Snack<theme>-Eat-Susan<agent>,” might also produce“Susan<actor>-Move Over” and “Cheese<theme>-Eat” (Goldin-Meadow 2003).Apparently, just as no child has to learn to factor experience into predicates andarguments, no child has to learn to use word order systematically to specify thesemantic role played by each element.

In sum, linguistically isolated children construct, out of their own thoughtsand communicative needs, systems that resemble the languages of the world inat least the following universal regards: all have words of more than one kind, atminimum nouns and verbs, organized into sentences expressing predicate-argument relations. The number of noun phrases is predictable from the meaningof the verb; the positioning of the nouns expresses their semantic roles relative tothe verb. Thus, the fundamental structure of the clause in both self-generated andmore established communication systems derives from the non-linguistic con-ceptual structures by which humans represent events, coupled with strong pref-erences for “flattening” these conceptual structures into linguistic expressions.This “cognitivist” interpretation of the origin of language in child conceptualstructure motivates all modern linguistic treatments of verb semantics that weknow of (Baker 2001; Chomsky 1981a; Dowty 1991; Fillmore 1968; Jackendoff1983; Rappaport-Hovav & Levin 1988). The same cognitivist approach figuresin most psychological theories about learning of both syntax and lexicon,whatever the other disagreements of their proponents (e.g. Gleitman, Gleitmanet al. 1988; Pinker 1989; Slobin 2001; Tomasello 2000). Indeed, both “nativist”and “learning-functionalist” wings of the language-learning investigative com-munity have seized upon the transparency and universality of such form-to-meaning correspondences in language acquisition as uniquely supporting their



learning positions in particular. (After the battle, the opposing generals retreatedto their tents to celebrate their victory.)

Young children use the structure of a sentence to guide interpretationof new verbs

Here we discuss evidence that young language learners, much like the infantlanguage inventors that we just discussed, exploit form-to-meaning correspon-dences as a rich source of evidence about the lexicon.

The case of argument number Particularly well studied has beenearly sensitivity to noun-phrase number as a cue to verb interpretation (Fisher1996, 2002; Naigles 1990; Lidz et al. 2003; Naigles & Kako 1993). For example,Naigles (1990) showed that children as young as 25 months of age interpret newverbs in accord with the number of their noun-phrase arguments. The childrenwatched a video-taped event in which two actions occurred simultaneously: inone composite display, a bunny pushed a duck into a bending posture, whilethe bunny and duck bent their free arms at the elbow. Each child heard thisdisplay described by either a transitive (“The bunny is gorping the duck!”) oran intransitive sentence (“The bunny and the duck are gorping!”). Following thistraining, the two subevents of the composite scene were shown separately ontwo side-by-side monitors, and the children were exhorted to “Find gorping!”One screen showed the causal event in which the bunny bent the duck; the othershowed the non-causal event in which both animals bent their arms. Childrenwho had heard the transitive training sentence looked longer at the causal event,while children who had heard the intransitive sentence looked longer at the non-causal event.

Similar syntactic evidence can persuade young children to alter their inter-pretation of a familiar verb. Naigles, Gleitman, and Gleitman (1992) askedpreschoolers to act out sentences using a toy Noah’s Ark and its associatedcharacters. The informative trials were those in which a verb was presented ina new syntactic environment, as in Noah brings to the ark or Noah goes theelephant to the ark. Young children adjusted the interpretation of the verb tofit its new syntactic frame, for example acting out go as “cause to go” (a.k.a.“bring”) when it was presented as a transitive verb.

Compare these results with the innovations of the deaf home signers whoinvented their own manual communication systems. In both cases, children mapparticipants in a conceptual representation of an event one-to-one onto nounarguments in sentences. Elsewhere we have proposed (Fisher 1996, 2000a;Gillette et al. 1999) that children might first arrive at this structure-sensitiveinterpretation of a sentence in a simple way – by aligning a representation ofa sentence with a structured conceptual representation of a relevant situation.



In this way a child might infer that a sentence with two noun arguments mustencode some conceptual relationship between the referents of the two nouns,while a sentence with only one noun argument might describe a state, property,or act of its single referent. This simple structure-mapping could take place assoon as the child learns to identify some nouns, and can represent them as partsof a larger utterance.

The centrality of argument number as a learning cue is further clarified byrecent studies that emphasize two important issues. One is that sensitivity toargument number makes its appearance astonishingly early in language acqui-sition, often before the child has uttered a single verb. The other is that thissensitivity can be demonstrated even in stripped-down experimental settingsthat remove all alternative evidence.

In a recent series of studies (Fisher 1996, 2002) children aged two, three, andfive years heard novel verbs in the context of (videotaped) unfamiliar causalevents; the verbs were presented either transitively or intransitively. The sen-tences contained only ambiguous pronouns, as in She’s pilking her over thereversus She’s pilking over there so that the sentences differed only in their numberof noun phrases. The children’s interpretations of the novel verbs were testedby asking them to point out, in a still picture of the event, which character’srole the verb described (Who’s pilking (her) over there?). Children at all threeages were more likely to select the causal agent in the event as the subject of thetransitive verb. Just as for the adult judges in the Gillette et al. (1999) studies,these findings provide evidence that the set of noun phrases in the sentence –even without information about which is the subject – influences young chil-dren’s interpretations of verbs. Recently these findings have been extended tochildren as young as 21 months of age. Fisher and Snedeker (2002) showed26- and 21-month-olds side-by-side videotaped events. One screen displayeda novel caused-motion event involving two people, and the other displayed anovel independent motion event involving only one person. As they watchedeach pair of scenes, the children heard either a transitive (He’s pilking him!) oran intransitive sentence (He’s pilking!). Children who heard the transitive sen-tence looked longer at the two-participant caused-motion event, while childrenwho heard the intransitive sentence tended to look equally at the two events(both of which displayed possible referents for an intransitive verb).

The findings reported so far are consistent with the view that there is abias to map one-to-one between the set of arguments of the verb and the setof participants in the event, in children acquiring an established language aswell as for linguistic isolates inventing their own sign systems. But perhaps,in the case of children learning an established language, the early honoringof this simple mapping from participant number to noun-phrase number is aneffect of language learning rather than the reflection of some unlearned bias.Do children simply exploit the most stable cues to mapping made available in



the language they hear, rather than relying on an unlearned bias for one-to-onemapping?

To investigate this issue, Lidz, Gleitman, and Gleitman (2003) askedpreschoolers to act out novel combinations of verbs and syntactic structuresin two languages: English (as in Naigles et al. 1992) and Kannada, a languagespoken in southwestern India. Kannada permits pervasive argument dropping,rendering the relationship between argument number and noun-phrase num-ber relatively variable in typical input sentences. Kannada also has, however,a causative morpheme that only occurs with causative verbs. The critical sen-tences pitted argument number (two nouns vs. one) against causative morphol-ogy (explicitly marked as causal or not). Kannada-speaking three-year-oldsignored the presence or absence of the causative morpheme, relying only onthe number of noun-phrases in the sentence they heard. In contrast, Kannada-speaking adults’ enactments were influenced by both morphology and argumentnumber. The adult findings again demonstrate that language learners ultimatelyacquire whichever cues to sentence meaning the exposure language makes avail-able. But strikingly, they also show that children are not totally open-minded:they appear to find some formal devices (argument number) more potent thanothers (inflectional morphology).

It is important to notice that the count-the-nouns procedure taken by itself iscoarse at best and fallible at worst. This is because nouns in the sentence andarguments of the verb are by no means the same thing. Often, for example, thereare too few nouns to match up with the event participants. This is for severalsystematic reasons, including incorporation phenomena of many kinds, and thepossibility of argument omission. In many languages, sentence subjects can beomitted if they are recoverable from context and prior discourse; in some lan-guages, including Chinese, Japanese and Korean, a verb’s direct objects can beomitted as well. Violations of any simple noun-counting principle are also obvi-ous in the reverse direction, for example when a language (like English) requiresa subject even for argumentless predicates (It is raining). And in any language,complex noun-phrases (John’s sister, a horse of a different color) contain morethan one noun, and sentences can contain adjunct phrases (with Ginger, in themorning), again yielding more noun phrases than argument positions.

Despite the complexity of the relationship between nouns in sentences andthe subcategorized arguments of the verb, several sources of evidence suggestthat ordinary sentences provide strong probabilistic information about the par-ticipant structures of verbs. For example, in the human simulations of Gilletteet al. (1999), adults benefited from simply being given an alphabetized list ofthe nouns in each sentence in which the mothers had produced a particularverb. In this case the adults (like the hypothetical learner) could not tell whichnouns were arguments of a verb and which were adjuncts, yet this linguistichint aided recovery of verb meanings from scenes. Li (1994; see also Lee &



Naigles 2002) analyzed speech to young children in Mandarin Chinese, andfound that although mothers often did omit noun phrases in sentences, mater-nal utterances still supported a systematic distinction among semantically andsyntactically distinct classes of verbs. Though arguments can be omitted, transi-tive verbs still occur with two nouns in the sentence more often than intransitiveverbs do, and systematically so.

Beyond argument number Experimental studies of novel verb learn-ing by young children have focused on argument number, in part because thisis an easily detectable cue to sentence interpretation. But clearly there’s morelinguistic evidence for lexical learning than argument number. Landau andGleitman (1985) argued, based on analysis of a blind child’s lexical develop-ment, that the child’s observation that look and see appeared in sentence com-plement structures made sense of her seemingly effortless acquisition of theperceptual nature of these verbs, while purely observational evidence yieldedno clear way to discriminate these from object-contact verbs like touch andhold. One can, for example, see that the sky is falling, and look how I’m doingthis, but not touch that the sky is falling. Argument type, like argument number,provides a powerful source of information for lexical learning.

Differences in argument type and number systematically map onto a semanticcross-classification of the verb lexicon, as revealed by naıve adults’ judgmentsof semantic relatedness among verbs (Fisher et al. 1991). Verbs that acceptsentences as their complements describe relations between their subjects andan event or state; these include verbs of cognition (know, think), perception(see, hear), and communication (explain, say). Verbs that take three noun-phrase arguments describe relations among the referents of those three nounphrases, typically transfer of position (put, drop), possession (give, take), orinformation (explain, argue). Later studies using the Fisher et al. proceduredocumented that these regularities could be recovered from a sample of Englishsentences produced in spontaneous child-directed speech in English (Lederer,Gleitman & Gleitman 1995) and Mandarin Chinese (Li 1994). Verbs’ syntacticbehavior, including both argument type and number, thus provides a sourceof information that systematically cross-classifies the set of verbs in much thesame way within and across languages, pointing to dimensions of semanticsimilarity. Indeed, it is this cross-classification – the set of structures associatedwith a single verb or small verb class – that accounted for subjects’ accuracyin the Jabberwocky (syntax-only) condition of the Gillette et al. (1999) humansimulation experiments.

Can young children, like these adults, profit from the full range of syntacticstructures they might observe with each verb? Considerable evidence tells usthat they are quite good at learning about the sentence structures in which partic-ular verbs occur (Gordon & Chafetz 1990; Snedeker, Thorpe & Trueswell 2001;



Tomasello 2000); such findings suggest that they may well be capable of takingadvantage of probabilistic evidence, presented across multiple sentences, forthe range of sentence structures assigned to each verb. Moreover, a computersimulation of syntactic learning from a sample of child-directed English (Brent1994) suggested that subcategorization frames for verbs could be recovered,based on very little prior syntactic knowledge (a few function words): an anal-ysis of verbs’ lexical contexts provided useful information for distinguishingamong verbs that are transitive or intransitive, or that take verbal or sententialcomplements (as in John likes to fish).

Beyond the first primitive mapping of two-noun sentences onto two-participant relations, mapping rules that are language-specific also come intoplay, further enriching the informational base and thus further increasing theefficiency and precision of predicate-term acquisition. The earliest-appearing ofthese is probably the interpretation of word order in multi-argument sentences:Hirsh-Pasek and Golinkoff (1996) reported that English-learning 17- to 19-month-olds were sensitive to word order in transitive sentences containingfamiliar verbs (Cookie Monster is tickling Big Bird vs. Big Bird is ticklingCookie Monster): they looked longer at a video screen on which the subjectof their test sentence was the agent of the target action. Children aged 21 and26 months show the same sensitivity to English word order when presented withmade-up verbs and unfamiliar actions (Fisher 2000b). Young children acquir-ing a free word-order language quickly acquire the semantic implications ofcase-marking morphology (e.g. results for Turkish learners reported in Slobin1982).

Children also develop more subtle language-specific expectations about themeanings of classes of words in their language; Slobin (2001) has termedthis phenomenon “typological bootstrapping.” For example, Talmy (1985)described systematic differences in the typical meanings of motion verbs inlanguages like Spanish (verb-framed languages) and English (satellite-framedlanguages). Spanish motion verbs tend to encode direction (enter, ascend),while in English, path information is relegated to a prepositional phrase, andverbs are more likely to encode manner (walk in, run up the hill). Native speak-ers of Spanish and English learn these tendencies, and develop slightly differentexpectations for the likely semantic content of a new verb (Naigles & Terrazas1998; for further discussion see Choi & Bowerman 1991; Fisher & Gleitman2002; Landau & Gleitman 1985).

The particular nouns that typically occur with each verb also undoubtedlyguide the child’s interpretation: drink and eat are not only transitive verbs;they systematically select animate subjects and different direct-object nouns(the potable and edible items). Data from “human simulations” suggest thatthis sort of information is helpful in both noun and verb learning (Gillette et al.1999; Kako & Gleitman in prep.; see Pinker 1989 for the initial statement of this



proposal). Quite young children have access to this sort of information as well:two-year-olds look at a picture of a glass of juice (rather than of a non-potableobject) when they hear the familiar verb drink (Fernald 2003), and successfullyinduce the referent of a new word introduced in an informative context as theobject of a familiar verb (e.g. She’s feeding the ferret!) (Goodman, McDonough& Brown 1998).

Summary

In the course of lexical development, children have opportunities to observeeach verb’s typical subcategorization frames, its typical nominal arguments,and the kinds of scenes or events that pertain when the verb is invoked. Althoughmuch work remains to be done to specify how each source of information isdetected and used by children, and how multiple sources of information interactin development, we argue that all of these sources of information converge tomake vocabulary learning efficient and nearly errorless.

Central to this so-called syntactic bootstrapping view is the interaction ofmultiple cues for word learning, trading off in different ways for differentclasses of verbs. This position does not replace observation of situations withlinguistic observations. Ultimately, word learning is a mapping problem. Thelearner must identify what (in the world, or at least in a human’s conceptionof the world) the surrounding community of speakers means by each word.Though our arguments focus on the ambiguity of referential settings – andthus the need for linguistic evidence to make vocabulary learning stable –the observations that give semantic content to words are observations of thenon-linguistic world. Sentence structures are relevant only to a subset of thedimensions of verb meaning, those that affect the number and type of argumentsassociated with the verb, and the temporal structure of the event it names (Fisheret al. 1991; Grimshaw 1990; Rappaport-Hovav & Levin 1988). In contrast, thevarious manners of motion encoded by slide, roll, and bounce have no directreflection in sentence structure (Fillmore 1968). When push comes to shove,only observations of the manner (and comparative violence) of actions in theworld will suffice to differentiate syntactically and semantically related verbs.

Lexical learning and the structure of linguistic knowledge

We have summarized evidence that, in order to acquire word meanings, childlearners amass all sorts of specific knowledge about individual words: theircontexts of use, their lexical-distributional properties (e.g. the association ofbake with cake) and the full range of their syntactic behavior. All this evidenceis required, in different degrees for different words, to converge on their mean-ings. The result is a knowledge representation in which detailed syntactic and



semantic information is linked at the level of the lexicon. What happens to allthis information, collected to solve the mapping problem? Is it disassembled,rather as a building scaffold is dismantled once the beams and bricks are inplace? We think not.

Experimentation on sentence comprehension in older children and adultssuggests the continued linking of linguistic distributional knowledge to partic-ular lexical items. Native speakers learn not only which sentence structures eachverb can grammatically combine with, but also how often each verb occurs ineach structure. Adults retrieve this information as soon as they identify a verb,and use it to bias online sentence interpretation (Garnsey et al. 1997; Trueswell& Kim 1998). Snedeker et al. (2001) demonstrated that both children and adultsresolved the ambiguity of such sentences as Tickle the frog with the feather andChoose the frog with the feather as a function of the frequency with which theseverbs naturally occur with noun-phrase versus verb-phrase modification. On-line parsing decisions by adults and by children as young as five are influencedby detailed and frequency-sensitive knowledge about the syntactic behavior ofeach verb.

These findings from the psycholinguistic literature mesh naturally with com-putational approaches to parsing that also represent syntactic representations asstrongly lexicalized: in Lexicalized Tree Adjoining Grammar, for example, thesyntactic possibilities of a language are represented by a finite set of tree struc-tures that are linked with individual lexical items, and a small set of operationsby which trees can be joined (Joshi & Srinivas 1994). This apparatus permits thestatement of syntactic dependencies (such as subcategorization) and semanticdependencies (such as selection restrictions), and yields a natural treatment ofnon-compositional idioms (kick the bucket). Such approaches are based on aclaim similar to the one we derive from examination of the learning procedure:an adequate description of the syntactic combinatorial principles of a languageis unworkable if kept separate from the lexicon and lexical learning. Similarly,independent evidence from crosslinguistically based theoretical linguistics sup-ports a view of language in which significant structural properties reside in thelexical component of the grammar (Borer 1984; Chomsky 1995c).

General and particular, and the requirementfor universal grammar

The manifest specificity of lexical organization that we have discussed through-out has often been taken as supporting a picture of language and its learning thatcan avoid appeals to unlearned constraints on the construction of grammars. Forexample, Tomasello (2000) and Goldberg (1995) have proposed construction-based accounts of language representation and acquisition, suggesting that chil-dren simply learn, word by word and construction by construction, how to



express each idea. According to this view, a more general and flexible grammargrows slowly from this piecemeal knowledge through general cognitive prin-ciples of induction and generalization. Proponents of such theories emphasizevariability and exceptions in the syntax–semantics mapping rules, and arguethat no constraints beyond those of the human cognitive/conceptual appara-tus are needed to account for the nature of human languages, and the facts oflanguage development.

However, we believe that this is a mistake. Although syntax must be rep-resented in the lexicon – to explain how we know, with such exquisite detail,which structures each verb appears in, and how we learn the meanings of theverbs in the first place – strong universal constraints on the alignment of syntaxand semantics are needed to explain the full set of facts. One issue, compellinglydiscussed by Mark Baker (2001), is the uniformity of clause-level structureswithin languages:

We do not find languages in which the verb meaning “hit” comes before the object,English-style (“The child might hit his parent”), and the verb meaning “kiss” comesafter it, Japanese-style (“The child might his parent kiss”). The word order of the objectand the verb is thus not learned purely by learning . . . individual verbs but must besomehow keyed into the process of learning the verbs as a class. (2001: 80)

If the organizational structure for predicate–argument structures could beanything at all, with the learner simply picking up these facts on an item-by-item basis, it would be hard to explain why each language organizes thesestructures so regularly, across verbs. The same is true for the basic semantic–syntactic linkages, even in languages at the extremes of linguistic diversity,as Baker shows by comparing languages such as English and Mohawk. Inall the languages of the world, though with occasional quirks and exceptions,not only do all the core participants in the action denoted by the verb getexpressed grammatically (the “theta criterion”), but the causal verbs put theiragents in subject position, the undergoers systematically surface as objects, thecomplements of mental verbs surface as clauses, and so forth. These linguisticproperties shared across cultures and language families, however otherwisediverse, imply strong restrictions on how we factor experience into predicatesand arguments, and what aspects of the conceptual predicates and argumentsare reflected in the organization of the clause. Children, being creatures like us,expect language to be organized in accord with such principles and thereforethey can learn it.


7 Empiricism and rationalism as research strategies

Norbert Hornstein

Introduction

A major influence on Chomsky’s approach to the study of mind has come fromrationalist philosophers such as Descartes. Like these thinkers, Chomsky’s workcan be usefully seen in opposition to empiricist approaches to mind articulatedby thinkers such as Locke.1 The aim of this chapter is to provide a conceptualbackdrop against which one can locate some of Chomsky’s claims. I try to dothis by outlining the different ways that the empiricist and rationalist traditions(actually, idealized versions of each) try to reconcile an apparent tension inepistemology (the theory of knowledge). The tension comes in trying to combinea theory of mind with a theory of truth to yield an account of how it is possible forhumans to know anything, i.e. have true beliefs in some domain, especially truebeliefs about the “outside world.”2 The two traditions endorse very differentconceptions of the relation of minds to the world. However, it is possible toconstruct a shared conception of what the epistemological project amounts tothat plausibly animates the details of the particular proposals that have beenadvanced. Doing this, I believe, allows for a better evaluation of the intuitionsthat drive these programs and thereby permits one to more fully appreciatesome of Chomsky’s main philosophical proposals. In what follows, I will tryto outline these general conceptions. I will then try to relate them to someof the concerns that Chomsky has raised in his linguistic and philosophicalwriting.

How you know

There are at least two kinds of questions that one can ask concerning the natureof knowledge: (a) What can be known? and (b) How does one come to knowanything? Often these questions arise together with the limits of knowledgecircumscribed on the basis of its mechanics. However, despite their links, thesecond question is largely independent of the first. In what follows, I focus onthe second. Answering it involves elaborating a theory of belief formation, atheory of truth, and an account of the interaction of the two.

145


146 Norbert Hornstein

More specifically, an epistemology is here taken as trying to account for howbeliefs arise, how they relate to each other, and how they relate to the world.How beliefs arise and relate to each other is the province of a psychology ofmind. How beliefs relate to the world concerns a theory of truth. Epistemologytries to combine the efforts of each of these two theories to provide a story thatexplains how it is possible to have true beliefs. Thus, for our purposes here,a theory of knowledge consists of three subparts: (i) a psychology or theoryof mind, (ii) a theory of truth, and (iii) a theory of how (i) and (ii) combine toallow some beliefs to be true. We shall see that the rather different psychologicaltheories developed within the empiricist and rationalist traditions can be seenas responding to the demands of the two different strands in an overall theoryof knowledge so conceived; empiricists responding to the exigencies of (ii) andrationalists reacting to the requirements of (i).

Before delving into details, let’s slightly elaborate each of (i)–(iii) above. Acore part of a theory of mind involves outlining how beliefs arise and interrelate.As such, it concerns itself with categorizing various kinds of beliefs, e.g. simpleversus general, abstract versus concrete, sensory versus cognitive, compositeversus elementary, etc. Second, a psychology tries to say how beliefs are gen-erated, e.g. they arise from sense experience, from operations on more basicideas, they are inborn, etc. Looked at in this way, a theory of mind requires apsychology which categorizes the varieties of our beliefs and outlines a set ofmental mechanisms and operations as a result of which our beliefs can be saidto arise and interconnect.

The second ingredient required for an epistemology is an account of truth.Such a theory says what truth is a property of. In particular, it tries to say whatit is for a belief to be true or for some idea to refer. Various kinds of theorieshave been explored in this regard. Two major variants have been realist theoriesof correspondence and non-realist quasi-verificationist approaches. The mostimportant property of the former is its radically non-epistemic nature. Thepicture animating the realist account is a sharp distinction between beliefs onthe one hand and the world on the other. Beliefs are true insofar as they “match”or “fit” the world. Unpacking what “fitting” or “matching” amounts to is animportant focus of such approaches.

It helps to understand what is meant by saying that truth is radically non-epistemic by contrasting realist theories with their non-realist counterparts.Non-realist approaches take truth as intimately connected to our epistemic abil-ities or our methods of inquiry. On this view, truth is the outcome, under somepossibly idealized conditions, of our investigative or cognitive capacities and/orprocedures. To say that some belief is true is to say that it is as highly valued asit can be along certain epistemic dimensions, e.g. coherence, naturalness etc.Outlining what these epistemic dimensions are and saying what sort of ideal-izations are appropriate is a central task of this approach. However, where this


Empiricism and rationalism as research strategies 147

view of truth contrasts with the former is in seeing truth as related to our cog-nitive powers in some (however attenuated) way. For the realist our cognitivepowers are what they are and the world is however it happens to be. There is noconceptual link between the former and the latter as there invariably is withinnon-realist conceptions.

The last part of a theory of knowledge elaborates the “fit” between the psy-chological theory of mind and the theory of truth. By mixing and matching thevarious accounts of mind with alternate theories of truth, one develops differentepistemologies and provides varying answers to the question of how true beliefis possible.

In what follows, three combinations will be discussed: (a) an empiricistpsychology coupled with a realist conception of truth, (b) a rationalist theory ofmind coupled with a realist conception of truth, and (c) a rationalist psychologyjoined with a non-realist view of truth. We will see that each combination hasits own charms. Hopefully, consideration of these views will also allow us tobetter understand some of Chomsky’s concerns.

The empiricist mind

Empiricist theories of mind (a version of Locke’s theory underlies what fol-lows) take beliefs to be complex entities made up of simpler parts, call them“ideas,” joined together by diverse mental operations. Ideas come in two vari-eties, simple and general. The latter are formed from the former by an operationof abstraction. Taking simple ideas as inputs, abstraction yields a general ideaby intersecting the properties of the input simple ideas. In this way, the prop-erties of general ideas are dependent on (and derived from) simple ideas. Notethat by being intersective, abstraction is a subtraction operation in the sensethat it deletes what is not common to the members of the group of simple ideasupon which the operation acts. Importantly, abstraction does not add anythingto properties of a general idea that is not part of the simple ones that yield it.

Simple ideas arise via sensation. They come in two varieties; those causedby primary qualities and those stimulated by secondary qualities. The formerconform to the contours of the stimulating quality and “resemble” them. Thelatter do not. Rather they are products of the mind and do not closely “resem-ble” those features of the external world that cause them to arise. A good partof empiricist reflections on the mind revolves around demarcating the borderbetween primary and secondary qualities as it is the relation of ideas to primaryqualities that is at the foundation of empiricist epistemology.

All simple primary ideas are formed (or fashioned or molded) in the mindby the action of the environment through the medium of the senses. The terms“formed,” “fashioned,” “molded” are meant to evoke the notion that the featuresof the simple idea are very closely related to the properties of the stimulus



situation. Simple ideas (at least those formed by primary qualities) directly (andaccurately) reflect features of the environment as transmitted via the senses. Theleading picture or metaphor is the mind as a wax tablet. The mind receives thesesimple ideas via sensation in much the way that hot wax receives the imprint ofan object impressed upon it. In the latter case, the contours of the imprint areclosely related to the physical properties of the imprinting object. In a similarway, empiricists urge, the properties of a simple idea in the mind can be traced tothe properties of the physical stimulus. In the case of minds, however, the “wax”is multi-modal and potentially able to receive inputs through each of the senses.A simple idea is consequently a multi-dimensional item, the inputs to each senseforming or molding one of the dimensions of the “idea.” Some of these inputsreflect the molding effects of primary qualities. One epistemological project isto figure out how to distinguish those simple ideas that are indeed “molded”by their causers and those that are not. This is not something that the minddoes naturally for us. It requires conscious effort and rigorous attention. Thissaid, however, once the ideas are appropriately categorized, there remains oneset, the simple primary ideas, that link minds to the world via a resemblancerelation grounded in a particular causal mechanism: for these ideas the externalstimulus molds the specific features of the idea that arises in the mind.

Observe that on this sort of account, simple primary ideas closely reflect thephysical properties of the environmental stimuli that produce them. They reflectthe environment because the medium on which the environment acts is receptiveand passive. With this view in hand, an account of mental contents reasonablyproceeds via an account of the structure of the environment and the stimuli itproduces. What is distinctive about empiricist theories is the emphasized causalconnection posited between the structure of our simple primary ideas and thatof the stimulus. With regard to simple primary ideas, minds are passive andwhatever structure they have is due to the molding effects of experience.

Behaviorism is a plausible outcome of empiricism so construed; for if aradical form of empiricist psychology is correct, then one gains little by con-centrating on the properties of minds rather than those of stimuli. Put anotherway, since the structure of (key) simple ideas is a reflection of the features ofthe environment and since it is much harder to examine minds than environ-ments it makes sense to concentrate one’s attention more on the structure ofstimuli, which are easily observed, than on minds, which are not. The crucialpoint about behaviorism is not its idiosyncratic disavowal of the mental but theview that key sectors of the mental are a mere reflection of the properties ofthe physical stimulus. In this sense, behaviorism is continuous with empiricistcharacterizations of the mind as passive and receptive.

Given this sort of empiricist view of how simple ideas arise, it is rela-tively clear in what abstraction consists. It is a second-order operation overthe imprinted properties of the simple ideas. In effect, it is a generalization



of the features of simple ideas. For this sort of procedure to work, there mustbe a relatively large data set to work on, a large number of simple ideas thatare fodder for the abstraction operation. The operation consists in taking whatis (sufficiently) common among the relevant simple ideas and dropping thedifferences.3 Note that if this is how general ideas arise, then the structure ofsuch an idea is also closely bound to the properties of the physical stimulus:the formation of the general idea is determined by the common properties ofthe individuals in the abstraction set and these individuals, it has been observed,are just simple ideas (some of) whose features are intimately connected to thoseof the physical stimulus. As such, the structure of some general ideas is alsoclosely circumscribed by the properties of the stimuli that give rise to the simpleideas.

Once we have general and simple ideas, various additional mental operationscan relate them to form beliefs. For example, predication can take a generalidea like “apple” and predicate to it the general idea “red” to form the belief“apples are red.” Other operations might combine general ideas to form yetmore complex general ideas, e.g. “Red apples” and more complex beliefs, e.g.negations of a belief, conjunctions of beliefs, etc. The intricacies of this processare not critical for current purposes. What is important is the close relationthat exists between the properties of the world (the stimuli) and the structure ofsome of our ideas and beliefs that an empiricist psychology of mental operationsadvances. Learning, the acquisition of ideas and beliefs, is a direct function ofthe environmental stimulus situation. A central concept is that of a formativestimulus and the leading picture of the mind is a blank, passive, highly receptivemulti-modal, repository of environmental input. We will return below to howthese psychological ideas can be exploited in an account of true belief.

The rationalist mind

Rationalist theories of mind markedly contrast with the one above. The maindifference lies in a firm rejection of the idea that experience exercises a shapingeffect on the structure of mental contents. Environmental stimuli, for ratio-nalists, are not stimuli that structure the contents of the mind. Rather, they aretriggers or occasions for the formation of ideas or beliefs. In other words, thoughexperience is causally necessary in explaining how beliefs arise in the mind itdoes not function the way it does within empiricist models to shape mentalcontent and structure. In particular, the idea that sensory input is (in some way)similar to the idea that it gives rise to (the way a wax impression is to the thingthat formed it) is largely rejected. There need be very little similarity betweensensory triggers and the ideas/beliefs to which they causally relate.

An analogy might help to make the rationalist view of mental operationsclearer. Ideas in the mind are like drawings in a dark, unlit room. Acquiring an



idea is like illuminating a picture in the dark. Experience is like flipping on alight switch. Before the light switch is flipped, the pictures are invisible. Thus,light-switch flipping is necessary to the pictures becoming visible. However,flipping on the light does not fashion or form the picture but allows one to seewhat is there quite independently. Nor is there anything very “similar” betweenwhat brings the pictures to visibility – the act of turning on the lights – and whatappears as a result of this action. The flipping of the light switch allows one tosee what is already there. It does not construct the pictures that are seen.

For rationalists, then, experience activates ideas that already inhabit the mind,it does not form them. Consequently, the relation between experience and thestructure of one’s beliefs and ideas, on this kind of account, can be very remote.As the analogy emphasizes, what appears as a result of turning on the light isa function of what pictures there are hanging on the wall, not on how one flipsthe switch.

Two features of this perspective are noteworthy. First, experience is requiredin accounting for the emergence of mental structures. To return to the analogy,the pictures remain invisible unless the light comes on. The rationalist whoposits rich innate mental structures need not (and did not) dispense with acausal role for experience. She or he just assigns a different role to experiencethan the empiricist does. Second, the role that rationalists assign to experienceis that of a triggering stimulus, a cause that activates ideas and beliefs whosestructures are (largely, if not wholly) what they are prior to the influence ofthe triggering stimuli.

Not surprisingly, this view of the role of experience dispenses with the ideathat the mind is like a blank wax tablet constructed by experience. Rationalistpsychology pictures the mind as a highly structured active entity. The minddoes not passively reflect experience but actively participates in interpreting it.Given a rationalist view of the role of experience in generating ideas, it is nolonger possible to be confident that what is “in” the mind (even “simple” ideas)is a faithful record of what is external to it (in the environment).

Note further that given this sort of view, general ideas need not be seen asarising via abstraction. The reason is that for a rationalist there is no reason forideas that are triggered to be closely related to the stimuli that trigger them.Therefore, it is possible to accept the notion that general ideas are (to a con-siderable extent) simply fixed natively in the mind. The same holds for beliefs.For a rationalist, it is unnecessary to distinguish simple ideas from general andideas from beliefs as regards origins. As a matter of historical fact, the innateideas rationalists proposed were by and large rather abstract. Ideas concerningparticulars arose from the interaction of general ideas and sensory experience.However, the details are irrelevant for present purposes. What is important isthe differences between rationalist and empiricist psychologies. The differencesreside in (a) the different causal roles each assigned to experience, formative



stimuli for empiricists, triggering stimuli for rationalists, (b) the different expla-nations of how beliefs arise, (c) the different views of how structured the mind isprior to experience and how active the mind is in manipulating stimuli, passive,blank and receptive or active, structured and participatory; and (d) the differ-ent attitudes towards “similarity” as characterizing the causal relation betweenenvironmental sensory input and the mental products resulting from sensoryexperience.

A reprise

In the presentation so far, I have outlined two different views of how minds arestructured. The views have been presented as incompatible. This is clearly mis-leading. The mind need not be exhaustively empiricist or rationalist in nature.A combination of mechanisms drawn from each tradition is possible and evenlikely. However, considering the non-mixed cases is instructive and I will con-tinue to contrast the two approaches as if they were exclusive.

One more point: in the presentation, I have heavily relied on analogies toconvey the differences between formative stimuli that structure ideas and trig-gering stimuli that merely reveal prior structure. The main contrasts can bemade without the analogies. What is conceptually central to empiricism is thecontention that the structure of the mind is closely related (“similar”), via oper-ations like imprinting and abstraction, to (at least some of) the structure of theenvironment. Thus, in accounting for the contents of belief, one must look tothe properties of the stimulus. The central feature of an empiricist approachis the kind of causal link – a formative one – postulated to exist between thecontents of thought and the structure of the environment.

Rationalists differ in substantially distancing mental structures from thecauses that allow them to become mentally active. There need be nothing sim-ilar between the cause and the ideas or beliefs that it gives rise to. As such, forrationalists, there is little of interest in psychologies based on abstractive andinductive mechanisms.

In sum, both empiricists and rationalists provide a role for experience. Theydiffer in what they see this role to be.

True belief

Before discussing the relative merits of these two approaches, let’s considerwhat account of knowledge (or true belief) we get by combining them with arealist conception of truth. So our first question is: given an empiricist psychol-ogy and a realist conception of truth, how is knowledge (true belief) possible?

Recall that for a realist, truth consists in correspondence between mentalobjects like beliefs and the non-mental properties of the world. As such, a



belief will be true only if the ideas which comprise it all denote/refer to entitiesin the world in the appropriate way. So, for example, the belief representedby the proposition “Apples are green” is true only if the property denoted by“green” is a subset of the properties which combine to form the general ideadenoted by “apple.” Further, on an empiricist account, this will hold just in caseall the individual apple-ideas have as part of their sensory imprint particularideas of green.

Recall that general ideas arise for an empiricist through abstraction over theset of particular ideas. The general idea can have no properties absent from eachof the particular ideas. As such, general ideas refer in virtue of the particularideas that they are related to via the abstraction operation. This feature is thekey ingredient in an empiricist psychology and it has an interesting property inthe context of an overall epistemology. It provides the outlines of a mechanical(causal) model that explains how people are able to successfully refer and holdtrue beliefs by explaining how it is that ideas can refer; in particular, how simple(primary) ideas refer and how via abstraction general ideas can as well. If beliefsare seen as combinations of such ideas, then it also provides an account of howit is that some beliefs (the ones formed in the right way) can be true.

This point is of some importance, so let me put it another way. For the empiri-cist, people are able to refer because they have ideas that refer. An empiricistpsychology explains how ideas can refer in the course of describing how theyarise. Recall that the properties of simple ideas faithfully track properties ofthe world. Experience shaping a passive receptive medium accounts for thepossibility of reference and truth. Ideas “latch onto” the world because theyare impressions of it. As simple (primary) ideas are analogous to wax impres-sions, they can refer to what is impressed upon them in much the same waythat a wax impression of a key can be used to identify or denote the key usedto make the wax impression. Similarity between impression and impressorhas quite a concrete meaning in cases such as this. For the empiricist, thisnotion is fully generalized, its chief virtue being a causally robust account ofreference.4

The empiricist account of general ideas as products of abstraction makestheir capacity for reference continuous with the account provided for simpleideas. Once again, how they arise causally accounts for how they can refer.Abstraction “adds” nothing not already contained in the simple ideas thatare fodder for the inductive process. Thus, abstraction does not alter simpleideas in ways that affect a general idea’s capacity to refer. Abstraction allowsgeneral ideas to retain their links to the world via the links that simple ideasestablish.

With this much in place, the rest of the story is conceptually clear. Beliefsare complex relations among ideas. They will be true only if the parts refer andrelate as the proposition expressing the belief indicates. Some of these relations



will result in true beliefs, some false. How the true beliefs can be true ultimatelyrests on the relations that simple ideas have to their formative stimuli.

The virtues of this sort of account are pretty clear. It provides a mechanical-causal model of true belief in its account of the aetiology of ideas. In a psycho-logical guise, empiricists, in effect, offer a causal theory of reference and truth.There are not many of these accounts to be had, so this sort of account doeshave its charms. The difficulties begin when one starts to seriously evaluate theplausibility of the learning theory.

Problems with empiricism

The main problem with empiricism resides in its psychology, in particular, thevery tight connection postulated to hold between the structure of our ideasand beliefs and the structure of the experiential input from the world. Theproblems arise when one considers in detail how an empiricist psychologymight account for the highly abstract, detailed, and structured knowledge thathumans have in some domains such as language and mathematics. The problemthat knowledge of language (or of geometry, a rationalist favorite) poses is thatin these domains the structure of our ideas and beliefs is not well describedas tracking (being similar to) stimulus input. The connection between beliefsattained and experience available to fix these beliefs is very remote. In fact, oncareful inspection, it appears that the kinds of experience relevant to formingthe requisite ideas/beliefs does not exist in the stimuli available to the individualwho acquires the relevant knowledge. The structures of many (if not most) ofour ideas and beliefs go far beyond the properties of the experience that we aresubject to. As such, the causal link between the environment and our beliefs,so vital to the inductive premises of empiricist psychology, is broken. If correctthis undermines empiricist epistemology by undermining its psychology. Foran illustration of the problem, consider the language case.5

Knowledge of language is a paradigm case of knowledge in the sense thata native speaker has complete and systematic mastery of a large body of verycomplex facts. So an interesting epistemological question is how it is that onecomes to acquire knowledge of one’s native language. The central fact aboutlanguage acquisition is that children are able to learn their native languagesrapidly and on the basis of degenerate, deficient, and inadequate data gatheredfrom the ambient environment. More precisely, children acquire a knowledgeof language despite the following inadequacies in the linguistic data set thatthey have access to.6

(a) The linguistic evidence a child has is imperfect; it includes slips of thetongue, incomplete thoughts, misstatements, etc.

(b) The knowledge of his or her native language that the native speaker attainsextends to an open-ended set of objects. There is no real upper bound on the



number of sentences native speakers can use and understand. This despitethe fact that the linguistic stimuli to which a child is exposed are merelyfinite. This implies that children must postulate rules on the basis of a limitednumber of example outputs of these rules. In sum, there is an inductive gapbetween what is attained (a rule) and the linguistic input to this acquisitionprocess (sentences/utterances conforming to this rule).

(c) There is considerable evidence that the knowledge attained by the nativespeaker includes features for which there is virtually no evidence in the datathat a child has access to in the course of acquisition.

Of these shortcomings (c) is the most problematic. It implies that no inductiveaccount can possibly explain knowledge of language, as there is no inductivebase on which to found such an account. Let me elaborate a bit as the point iscrucial.

Points (a) and (b) together can be taken as indicating difficulties for a tradi-tional empiricist approach to language. Recall that for an empiricist the relia-bility of our beliefs depends on the qualities of the environmental stimuli onwhich they are based. What (a) and (b) point out is that the data base may bemore misleading than one might initially suppose. However, (a) and (b) do notin themselves pose an insurmountable obstacle. What they point out is that oneneeds to be a little sophisticated in how one exploits the environmental input. Anempiricist psychology will have to statistically massage the inputs in order toovercome the hurdles posed by (a) and (b). (a) poses a standard kind of samplingproblem. So long as the linguistic input is not systematically misleading it willbe possible to use the data to overcome the problem of imperfect “sentences”that (a) notes.

(b) may prove to be more challenging. What is attained are rules. The evi-dence are bits of spoken/produced language. The problem is to figure outhow to generalize from instances of a rule to the rule. This is not an entirelytrivial problem but it is the sort of problem that empiricists are comfortableaddressing.7

Neither (a) nor (b) force us to deny that our ideas and beliefs about lan-guage are formed by experience. (c) poses a much more significant problem.It claims that for large parts of a native speaker’s attested linguistic knowledgethere is no relevant experience available at all on the basis of which the rele-vant ideas/beliefs could be induced. If this is so, then not even a sophisticatedempiricist theory can suffice, for there is no way to fashion an inductive linkbetween our ideas/beliefs and any relevant linguistic input. This points to theconclusion that some of a native speaker’s linguistic knowledge is already inthe mind – it is innate – and is not formed/structured by environmental inputs.Note, once again, that this does not imply that experience is irrelevant to ourcoming to have this knowledge. Instead, that experience triggers, rather thanforms, the relevant mental structures that underlie this knowledge.



Note that once innate beliefs are needed, they can be used to mitigate theproblems raised by (a) and (b) above. We can evaluate inputs against the expec-tations that the innate beliefs give rise to. We can use our innate beliefs to helpget us to rules whose general features are innately provided from instances ofthese rules. Once innate beliefs are required, they serve to simplify the inductiveproblem quite generally.

The argument above, the argument from the poverty of the stimulus, is astaple of the rationalist argument against empiricist approaches to knowledge.In more classical rationalist texts the argument is made by considering mathe-matical rather than linguistic knowledge. However, the point is the same; whatwe know about numbers, or geometrical figures or sequences, etc. far outstripsthe relevant inputs. There is an unbridgeable gap between what we know andwhatever evidence we could have used to come to know it. One of the contri-butions that Chomsky has made to the rationalist position comes in providingone very well worked-out and detailed example of where knowledge attainedoutruns inputs available. Given that linguistic knowledge is both a mundane anda paradigm case of knowledge (what better example is there of knowledge thanknowledge of one’s native language?), the failure of empiricist psychology inthe linguistic case suggests that the empiricist epistemological project is gen-erally inadequate. Moreover, it suggests that once one begins to look carefullyat specific domains of knowledge, it quickly becomes evident that the poverty-of-the-stimulus problem characteristic of the linguistic and mathematical casesgeneralizes all too easily.

An epistemological consequence

The rationalist critique of empiricist psychology has epistemological repercus-sions. Recall that what makes an empiricist theory of mind attractive is that itnicely combines with a realist conception of truth to say how it is that ideascan refer and beliefs can be true. The problem is that the psychology requiredto make this account run is simply inadequate. The question that arises is whathappens if we substitute a rationalist approach to the mental. Do we retain asimilarly satisfying account of how beliefs can be true?

Not really. Recall that it is the wax tablet model that undergirds the causalaccount of reference in an empiricist epistemology. The rationalist rejects thispsychological model. This makes it impossible for the rationalist to exploit theempiricist’s strategy of accounting for how true belief and reference is possibleby observing how it is that ideas and beliefs are fixed in the mind. For therationalist, how beliefs arise is not bound in the right way to the structure ofthe environment. There is nothing “similar” between stimulus and idea/belief.Thus, there is no way to ground the reference relation in the mechanics ofidea/belief fixation. The rationalist’s notion of experience as a trigger, and the



mind as active and structured, prevents him or her from advancing an accountof truth and reference grounded in the aetiology of ideas and beliefs as therationalist account crucially exploits notions of mind at odds with one neededto get such an account off the ground. As such, the rationalist cannot import theempiricist’s account of how true belief and successful reference is possible.

What account does the rationalist provide? There is no real account; onlythe blunt statement that we are “built” for (some) truth. As Descartes put it, wecan have true beliefs because a benevolent God implanted in us some ideas thatrefer and some beliefs that correspond to reality. Put more bluntly, that some ofour ideas refer and some of our beliefs are true is something of a miracle.

There is no “theory” here, just an assertion. Moreover, the anemic nature ofthe account of true belief can be traced to what is arguably the principal virtueof the rationalist story: the sophisticated nativist psychology. By distancingthe aetiology of the structure of ideas and beliefs from the stimuli that promptthem the rationalist cannot then easily turn and use the causal link afforded byenvironmental stimuli for epistemological purposes.

This feature of rationalist theories may be useful in accounting for a curioussociological fact among philosophers. Empiricism seems to be the default viewamong philosophers. A general distrust of rationalist claims is quite standard.Part of this may stem from an appreciation of the tension noted here. If oneholds a realist conception of truth then it might appear that the only way toavoid a “miracle” theory of knowledge is to adopt an empiricist psychology.

Theories of truth

The above illustrates an interesting tension within an overall epistemology. Onthe one hand we have empiricism which affords a mechanical account of howtrue belief and reference is possible, but it does this by adopting a suspectpsychological theory. On the other hand, there is rationalism which provides apromising view of the mental but has little to say about how people are able touse ideas to refer and how they can have true beliefs.

A possible resolution to this tension comes from considering the possibilitythat the problem lies with the theory of truth heretofore held constant. Recallthat there are three parts to an overall epistemology: a psychology, a theory oftruth, and an account that explains how ideas and beliefs as described by thepsychology can refer and be true in the sense offered by the theory of truth. Tothis point we have considered how to combine two psychologies with a realistconception of truth and found the mixes considered wanting. Perhaps we couldavoid these troubles by combining a rationalist theory of mind with a non-realisttheory of truth.

Recall that non-realist conceptions of truth allow truth to be tied to ourcognitive capacities and investigative procedures. Truth (and reference) on such



a view is what our exercised capacities yield, perhaps in the limit. Whether thisview can move from metaphor to proposal is unclear. However, the intuitionthat animates it is tolerably clear. Truth is linked to our epistemic capacitiesand is not radically distinct from them, as the realist maintains. The promise ofthis view of truth is that it might be combinable with a rationalist view of themind without requiring a “miracle” view of truth and reference. Let me assignsome (very!) tendentious tags to the two combinations we are considering. Calla rationalist psychology combined with a realist theory of truth a “Cartesian”view. Call a rationalist psychology combined with a non-realist view of truth a“Kantian” view.

The central intuition animating the Cartesian is that our capacities for knowl-edge are in some sense a fortuitous accident. It arises from a brute fact withonly God’s benevolence to explain it; namely, we are built for truth, at least insome domains of inquiry.8 The Kantian alternative reverses the slogan: we arenot built for truth so much as truth is built by us. It is the output of our capacities(again under perhaps idealized circumstances) and so it is not surprising thatwe are able to have true beliefs and to refer successfully as this comes down tolittle more than saying that there are good beliefs versus bad beliefs as judgedby our own lights. The good ones are the true ones. After all, what would itmean to say that some belief or other is as good as one might hope (the simplest,most elegant, most explanatory, perfectly predictive and retrodictive, etc.) butis nonetheless false? The Kantian scheme might yield a non-trivial account oftrue belief by elaborating the cognitive parameters along which beliefs are tobe evaluated. Truth will be epistemic superiority along these dimensions.

Some Chomskyan reflections

It is not my aim here to evaluate the alternatives surveyed above. Rather I wouldlike to use the general background issues outlined above to highlight some ofChomsky’s concerns.

Chomsky is clearly partial to the rationalist critique of empiricist learningtheory. As noted above, his own work in grammatical theory constitutes oneof the best-worked-out cases in favor of a rationalist approach to the mind. Itis clear from the case of linguistic knowledge that the mind is already highlystructured before its encounter with linguistic experience. Though linguisticexperience is important for language acquisition, it plays its role in the contextof rich mental structures that form part of human biological endowment. Thisview clearly resonates with rationalist beliefs.

A reflection of this rationalist commitment can be seen in Chomsky’s distinc-tion between I-language and E-language. The former is the individual, internal,intensional system of knowledge that undergirds a person’s linguistic capacities



(Chomsky 2000a: 5). The latter marks the sense in which Dutch and Germanare different languages (Chomsky 2000a: 48).

Chomsky has argued that the latter notion is of little use if one’s interest is ascientific theory of language. The E-language notion has never been well definedand has played virtually no role in any naturalistic approach to explaininglinguistic phenomena. Despite this it has great staying power. Why so? It is(at least in part) a reflection of empiricist commitments that many find hard toshake off. Consider how these empiricist commitments require a construct suchas E-language.

Given an empiricist psychology, one’s internal states reflect the properties ofthe environment that one is exposed to. It is clear that people attain linguisticcompetence. As such, they must have learned their native languages. But, givenempiricist views of learning, this means that a native speaker’s psychologicalstate must have tracked some mind-external properties. Which? Those of theE-language. The picture that emerges is something as follows (cf. Chomskyparaphrasing Dummett, Chomsky 2000a: 48):

[Language] is a particular social practice “in which people engage,” a practice that“is learned from others . . . ” [Language] . . . exists “independently of any particularspeakers”; every individual speaker “has” such a language, but typically has only a“partial, and partially erroneous, grasp of the language.”

Typically, this learning is characterized in crudely behaviorist terms. Quine’sdiscussion of this process in Word and Object serves to illustrate this:

The child’s early learning of a verbal response depends on society’s reinforcement ofthe response in association with the stimulation that merits the response, from society’spoint of view, and society’s discouragement of it otherwise. (1960: 82)

In short, the need for E-languages seems to be just the sort of object requiredonce one adopts an empiricist approach to acquisition.

The problem, as Chomsky notes, is that there is “no useful general sensein which we can characterize “language” so that Dutch and German are twodistinct “languages,” which people know only “partially” and “erroneously.”Moreover, the lack of such a general notion is a “commonplace” of empiricalwork within linguistics (Chomsky 2000a: 48). Furthermore, as Chomsky hasargued in great detail, the general picture of learning that requires notions likeE-language has very little to recommend it empirically.

In short, the current debate over which notion of language to adopt for nat-uralistic investigation, I-language versus E-language, reflects the themes thatanimated the earlier divergent empiricist and rationalist approaches to mind.

Chomsky believes that what holds for linguistic competence holds much moregenerally. It is clear that in many domains our attained capacities far outstrip thedata that cause them to arise. The poverty of the stimulus is easily detected in



almost every domain. This holds for various perceptual domains such as visionand hearing and cognitive domains such as our concept of number, cause, andperceptual body. It most likely also extends to our science-forming capacitieswhere, it has been regularly noted, theory also far surpasses the inputs/data thatundergird it. Though we know very little about the science-forming “module”of the mind, there is little reason to resist the conclusion that it too is partiallybased on innate features given the under-determination of its outputs (theories)by its inputs (data in some domain) (Chomsky 2000a: 82).

Chomsky suggests that part of the success of empiricism lies with the termi-nology used to discuss the relevant issues. He suggests that we dispense withthe notion that knowledge development is due to learning. Rather, knowledge“grows”/“matures” in us much the way that bones do or the way that pubertydoes. Acquisition of knowledge on this suggestion is not something that we do,rather it is something that happens to us and whose course of development islargely due to the nature of our internal (biological) constitutions.

This switch in metaphors clarifies the rationalist conception of the mind. Itrefocuses the role of the environment in the development of various competen-cies that arise in us by highlighting the inner dynamics of an unfolding nativeconstitution. Note that the growth metaphor leaves intact important causal linksto the environment. One does not grow without the right nutrients. However,there is no temptation when thinking of growth to downplay the structure of thegrower in charting the course of development, as there seems to be when onethinks of acquisition in terms of learning instead of growth.

Chomsky suggests a second area for terminological revision: dump the ideathat linguistic competence is a kind of “ability.” As he notes, once one seeslanguage in E-language terms as a social practice whose development in thechild is tied to the shaping effects of community and environment, “it is tempt-ing to understand knowledge of language as the learned ability to engage insuch practices . . . ” (Chomsky 2000a: 48). The “ability” view contrasts with theconception of language as a

generative procedure that assigns structural descriptions to linguistic expressions, knowl-edge of language being the internal representation of such a procedure in the brain (inthe mind, as we may say when speaking about the brain at a certain level of abstraction).From this point of view, ability to use one’s language (to put one’s knowledge to use) issharply distinguished from having such knowledge.

Once again, the suggested terminological change rests on the inadequacy ofthe empiricist theory of the mental that lurks behind the term “ability.” Empiri-cists, for the reasons outlined earlier, are partial to concepts that reduce thedistance between a manifest capacity and the shaping effects of the envi-ronment that are presumed to underlie it. Notions such as ability, behavior,E-language, and learning all carry empiricist resonances. To overcome the subtle



empiricist influence the use of such terms can have, Chomsky urges that wereplace them with others more congenial to the rationalist picture of our mentallife.

A second area which reflects Chomsky’s rationalist concerns relates to hisskepticism concerning how the epistemological project outlined in earlier sec-tions has been conceived. The project outlined above revolves around the ques-tion of how beliefs can be true and ideas refer. Chomsky (2000a: 148–50)suggests that this is not a well-formulated question as it is not ideas that refer(at least in ordinary usage) but people who refer by using ideas in various(context-dependent) ways.

The question “to what does the word X refer?” has no clear sense, whether posed forPeter, or (more mysteriously) for some “common language.” In general a word, even ofthe simplest kind, does not pick out an entity of the world, or of our “belief space”. . . .(Chomsky 2000a: 181).

In ordinary usage, . . . person X refers to Y by expression E under circumstances C, sothe relation is at least tetradic; and Y need not be a real object in the world or regardedthat way by X. More generally, person X uses expression E with its intrinsic semanticproperties to talk about the world from certain intricate perspectives focusing attentionon particular aspects of it, under circumstances C . . . (Chomsky 2000a: 150)

Similarly, it is not beliefs that are true but assertions made by individuals underspecific circumstances that employ beliefs. The problem then is not how to setup a correspondence between ideas and objects but to explain particular kindsof language use. Chomsky explicitly notes the rationalist flavor of this view ofthe semantic enterprise. Quoting from Cudworth, he summarizes a rationalistposition with which he sympathizes:

An approach to semantic interpretation in such terms has a traditional flavor. Seventeenth-century rationalist psychology held that innate “cognoscitive powers” enable people “tounderstand or judge of what is received by the sense,” which only gives the mind “anoccasion to exercise its own activity” to construct “intelligible ideas and conceptionsof things from within itself” as “rules,” “patterns,” “exemplars,” and “anticipations”that provide relations of cause and effect, whole and part, symmetry and propostion,characteristic use . . . unity of objects and other Gestalt properties, and in general “onecomprehensive name of the whole.” (2000a: 181ff.)

As the above quote makes clear, Chomsky, following the observations of earl-ier rationalist thinkers (and some modern philosophers like Goodman), believesthat it is unlikely that there will be a mechanical account of such “referential”activity, contrary to what the empiricist vision might suggest. What counts ashaving successfully referred is context dependent and subject to a variety ofidiosyncratic concerns and interests. It would be surprising if anything verygeneral unified all these instances of reference once one gets beyond the cog-nitive resources that these various referential uses exploit. The same goes for



asserting truly. If this is correct, then the entire project of trying to ground epis-temology in some sort of mechanical causal theory that animates the empiricistprogram is simply misconceived.

An expression such as “I painted my house brown” is accessed by performance systemsthat interpret it . . . and articulate it while typically using it for one or another speechact, on the productive side. How is this done? The articulatory-perceptual aspects . . . arestill poorly understood. At the conceptual-intentional interface the problems are evenmore obscure, and may well fall beyond human naturalistic inquiry in crucial respects.(Chomsky 2000a: 125)

Observe, if Chomsky’s arguments are even roughly correct then this removesthe main difficulty with the rationalist alternative to empiricism noted above,viz. that it provides no compelling account of how beliefs can be true and ideasrefer. Put another way, the main virtue of the empiricist approach was its promiseof a non-trivial account of true belief and reference. If this promise cannot befulfilled, then there is little left to recommend the empiricist perspective.

A third area where Chomsky’s rationalism appears is in his discussion ofproblems ripe for naturalist inquiry. Chomsky draws an important distinctionbetween problems and mysteries. He has observed that there are some areas thatyield to naturalistic inquiry and many, the vast majority, that do not. The formerare filled with problems, questions that can be attacked and understood usingthe procedures of scientific inquiry. In other domains the standard techniques ofscientific inquiry gain no traction and all we are left with are pseudo-questionswhose potential answers are as mysterious as they have ever been.

Among the aspects of the mind are those that enter into naturalistic inquiry; call themthe “science-forming faculty” (SFF). Equipped with SFF, people confront “problemsituations,” consisting of certain cognitive states . . . questions that are posed, and soon . . . Often SFF yields only a blank stare. Sometimes it provides ideas about how thequestions might be answered or reformulated, or the cognitive state modified, ideas thatcan then be evaluated in ways that SFF offers (empirical test, consistency with otherparts of science, criteria of intelligibility and elegance etc.). (Chomsky 2000a: 82–3)

Chomsky notes that we should expect that some domains of inquiry willprove efficacious and some not. After all, if the mind is richly structured, thenthere is reason to expect that this structure would impose limits on the domainsin which naturalistic inquiry can be successfully pursued. Given the rationalistview that what underlies our deepest and most complex capacities is a highlystructured mind, it is reasonable to think that there will be intrinsic limits towhat we can know and that therefore there may well be some domains withinwhich naturalistic inquiry will never gain a real foothold.

Like other biological systems, SFF has its potential scope and limits; we maydistinguish problems that fall within its limits and mysteries that do not. (Chomsky2000a: 83)



There may well be many . . . questions that are not subject to empirical inquiry in themanner of the sciences . . . if humans are themselves part of the natural world, and thushave specific biological capacities with their scope and limits, like every other organism.(Chomsky 2000a: 73)

This idea resonates with the Cartesian view outlined above. Recall that thecrucial intuition behind the Cartesian view is that knowledge is a fortuitousaccident. We can gain significant knowledge only if our minds are built fortruth in one or another domain. Only then is deep insight possible. However, ifthe overlap between mind and the world is accidental, then there is no reasonto believe that it will be generally applicable to any question that might interestus or even be of importance to us. Limits to knowledge are inherent given thispicture of things.

The successful natural sciences . . . fall within the intersection of SFF and the nature ofthe world; they treat the (scattered and limited) aspects of the world that we can grasp andcomprehend by naturalistic inquiry, in principle. The intersection is a chance product ofhuman nature. (Chomsky 2000a: 83)

This contrasts with both the empiricist view of matters and the “Kantian”perspective briefly parodied above. For the former, the mind is an all-purposedevice whose structure is a function of what inputs it receives. The mind, beingblank, is not predisposed to inquiry in any direction. One domain is neitherbetter nor worse than any other. Thus, the distinction between problems on theone hand and mysteries on the other finds no place.

The “Kantian” retains some of the empiricist’s viewpoint. For him or hertruth is a function of our own capacities and methods of inquiry. As such, itwould appear that the possibility of radical failure should be hard to formulate.Mysteries are temporary. With time they should convert to problems. The ideathat some problems are intrinsically unsolvable for us, an idea that the Cartesianemphasizes as a real possibility, fits ill with the idea that truth is, in the end, afunction of us.

Chomsky (1988b) observes that if humans are no different from other animalsthen there should be some things that they will do well as they are constructedto succeed in these domains, and some that they will fail at because of theirinherent cognitive limitations. As he notes, the failures are the price we pay forour successes. Without a richly structured mind we would be unable to advancetheories that far outstrip the data that suggest them. As all interesting theoriesare vastly underdetermined by the evidence that support them, it would seemthat our successful explanations rest, in part, on the resources of our structuredminds. However, with all advantages come limitations. If it is indeed the casethat our successes are spurred in part by innate gifts of nature, then it is naturalto assume that these same structures should act to limit what we can come toknow. This has a clear Cartesian resonance.



Conclusion

Chomsky has quite self-consciously located his work in a rationalist setting.Like earlier thinkers, he has stressed the distance between our knowledge andour behavior, between our environment and our mental states, between what weknow and how we put our knowledge to use. In a contemporary setting, theserationalist commitments set Chomsky apart from many thinkers still animatedby empiricist concerns. The aim of this essay has been to illuminate some ofChomsky’s concerns by locating them against the more general backdrop ofempiricist–rationalist differences.


8 Innate ideas

Paul Pietroski and Stephen Crain

Here’s one way this chapter could go. After defining the terms “innate” and“idea,” we say whether Chomsky thinks any ideas are innate – and if so, whichones. Unfortunately, we don’t have any theoretically interesting definitions tooffer; and, so far as we know, Chomsky has never said that any ideas areinnate. Since saying that would make for a very short chapter, we propose todo something else. Our aim is to locate Chomsky, as he locates himself, in arationalist tradition where talk of innate ideas has often been used to expressthe following view: the general character of human thought is due largely tohuman nature.

One can endorse this view without saying that humans have specific concepts,like turnip or carburetor , independent of experience. Correlatively, itis important to remember that while Chomsky is a nativist about language,he does not think that specific languages are innate. Whether a child ends upspeaking Japanese or English, or both, clearly depends on the child’s experience.The nativist claim is that all natural languages share core features that reflectthe biology of homo sapiens. Knowing a language, like having a heart, is areflection of our biological endowment. Just as humans have internal organs withcharacteristic traits, they speak languages with characteristic traits. Accordingto Chomsky, linguistic variation is severely constrained by the mental systemsthat make human language possible. But this is compatible with some linguisticvariation across (and within) communities, since the traits of individuals arealways products of nature and nurture. Diet and exercise affect one’s heart,within a limited range of possibilities. Similarly, experience affects the courseof a child’s linguistic development, within certain parameters.1

Thus one can be a nativist about language without saying that humanscome into the world equipped with particular languages, like cars come offthe assembly line equipped with wheels. Similarly, one can be a nativist aboutideas without saying that humans come into the world equipped with particu-lar ideas. Encounters with the world clearly have an impact on the ideas thathumans acquire. Nevertheless, biology may well impose substantial constraintson the “space of possible ideas” that are naturally available to humans, much as

164


Innate ideas 165

biology constrains the “space of possible languages” that are naturally avail-able to humans. Moreover, experience can have an impact without “shaping”the ideas that humans naturally employ. If children do not learn languages bygeneralizing from experience, and a child’s history of linguistic stimulationsimply drives the child down one of the biologically available linguistic paths,then perhaps the same is true of the acquisition of mental capacities more gen-erally. A nativist perspective is reinforced in so far as mental capacities emergein young children throughout the species, in settings where experience dramat-ically underdetermines the knowledge that children attain. As with language,perhaps experience plays the role of triggering innate mental resources in theformation of (at least many of) the ideas that humans naturally employ – ideasthat can develop only in a limited number of ways. If so, while experienceinfluences the particular course of idea-formation in any given individual, thisis not (as empiricists suggest) because individuals seek and find regularities intheir experience. Rather, regularities are imposed on experience in accordancewith mental structures already in place.

We think this is Chomsky’s view, and also the view he finds in certain histori-cal figures who participated in debates about innate ideas. Chomsky’s contribu-tion to the traditional debate lies in (i) his articulation and defense of a detailednativist program in linguistics, showing how experience plays only a restrictedrole in a broadly rationalist account of the acquisition of linguistic knowledge,and (ii) the framework this program suggests, given its empirical success, forthe more general study of human cognition. Linguistics – where this includesnot just the study of expressions and their properties, but also related work inpsycholinguistics – provides a case study of how to investigate which aspects ofhuman thought are due largely to human nature. Earlier chapters have addressed(i). We’ll try to give the flavor of (ii) by discussing some historically impor-tant examples, and then by reviewing some recent discoveries, inspired by theChomskyan approach to human psychology, about the properties of linguisticexpressions that have a direct bearing on logical reasoning.

Experience, mind, the gap

As the previous chapter indicated, Chomsky is part of a rationalist traditionaccording to which human knowledge is rooted in the cognitive resources usedwhen humans conceptualize their experience. Experience can, and often does,“trigger” the use of these resources. So humans deprived of normal experiencemay not develop in the usual way, and they may be unable to apply theircognitive resources in the usual knowledge-producing ways. But rationalistsmaintain that in at least many domains, the knowledge we do achieve goesfar beyond our experience in ways that reflect the contours of our cognitive


166 Paul Pietroski and Stephen Crain

apparatus. From this perspective, the knowledge achieved is not the result of“generalizing” from experience: the character of human knowledge owes moreto our shared human nature than to our shared human experiences. Knowledgeof language is a paradigm example of a domain in which such “poverty-of-stimulus” considerations strongly favor a rationalist approach; see the previouschapter.

Although our cognitive systems surely reflect our experience in some manner, a carefulspecification of the properties of these systems on the one hand, and of the experiencethat somehow led to their formation on the other, shows that the two are separatedby a considerable gap, in fact, a chasm . . . . The problem, then, is to determine theinnate endowment that serves to bridge the gap between experience and knowledgeattained. . . . The study of language is particularly interesting in this regard. (Chomsky1986)

As this quote makes clear, Chomsky sees his “rationalist linguistics” as onebranch of a broader – and for the most part, yet to be developed – rational-ist psychology; see especially Chomsky (1966, 2002a). A similar programwas envisioned by various seventeenth- and eighteenth-century theorists whomChomsky regularly cites as initiators of the (very much unfinished) cognitiverevolution: Descartes, the Port-Royal logicians, Ralph Cudworth, and Herbertof Cherbury. These thinkers maintained that a great deal of human knowledgesprings (under the pressure of normal experience) from cognitive resourcesthat humans have independent of and prior to experience. In their terminology,human beings have innate ideas.

Having introduced this traditional terminology, some caveats are in order.The claim is not (and was not) that humans are born with certain ideas – say,the idea of a triangle, or the idea of a verb phrase – and then just wait for thechance to apply them. Whether humans are born with ideas is not important on arationalist account. The relevant cognitive resources may develop according to amaturational timetable, or they may be triggered by experiences that occur laterin life.2 Moreover, a rationalist need not describe any pre-triggered cognitiveresource as, say, an innate idea of a triangle. Perhaps a person can be properlysaid to have an idea of X only if the person has had triggering experiences thatare “appropriate to” the formation of that idea. If so, then nothing within usprior to experience can be properly described as an idea of X, no matter whatwe substitute for “X.” Rationalists can agree to limit use of the word “idea” inthis way. For this is fully compatible with the hypothesis that humans come tohave the idea of a triangle (or a verb phrase), across a range of environmentalsituations, mainly because human cognition has contours that make us apt toacquire such ideas. Indeed, a rationalist need not describe any pre-triggeredcognitive resource as an idea. Perhaps every idea is an idea of something, and


Innate ideas 167

so a thinker has an idea only if it is an idea of X, for some particular instanceof “X.” On this view, all ideas are intentional; they have a specific content. Arationalist could adopt this view while granting that mental entities come tohave specific contents only under the pressure of experience.3

The traditional terminology is thus misleading. Innate ideas need not beinborn; experience is relevant; and whatever is innate need not be an idea,much less an idea of anything in particular. We suspect this is why Chomskyaligns himself with Descartes and other rationalists, but without using the phrase“innate ideas” to describe his own views. For Chomsky, the important issuesdo not concern what counts as an idea. The important issues concern (the sizeof) the “gap” between experience and knowledge, and how to best characterizethe cognitive resources that fill the gap – thereby making it invisible in ordinarylife.

In hindsight, it seems clear that many empiricists have rejected rationalistproposals at least in part because of misunderstandings about the crucial featuresof those proposals. To the best of our knowledge, no empiricist has successfullyrebutted the relevant poverty-of-stimulus arguments for a broadly rationalistpsychology.4 One can sympathize with the desire (discussed in the previouschapter) for an account of how humans occasionally come to have veridicaland justifiable beliefs about the mind-independent world. But one must guardagainst letting this desire drive one’s psychological theorizing. According toDescartes and Chomsky, questions about how minds are related to the worldare hard questions to be faced honestly in light of (i) our best theories of humanpsychology and (ii) our best theories of the mind-independent world. If ourbest theories of human psychology turn out to be rationalist theories, then sobe it, even if this makes it really hard to see how we ever manage to come tohave true justified beliefs about the mind-independent world. One can’t arguefor an empiricist psychology by noting that if human knowledge were mainlya product of generalizing from our shared experiences, then it would be easierto account for empirical knowledge.5

Why it’s called “Plato’s Problem”

It might help at this point to focus the discussion on a historically importantexample, already mentioned, that reveals how far human knowledge can extendbeyond experience.

If a triangle is a three-sided figure with perfectly straight lines, then no onehas ever seen a triangle, at least not in the ordinary sense that we can anddo see rocks, people, and chalk marks. Nor has anyone ever heard, touched,smelled, or tasted a triangle. Nonetheless, humans have the idea of a triangle;and one can know a great many things about triangles. For example, the ancient



Greeks – who also had the idea of a square, a right triangle, the area of a figure,and the sum of two areas – knew that the square built on the hypotenuse ofa right triangle with equal sides must be equal to the sum of the areas of the(equal) squares built on the shorter sides of the triangle. These ideas correspondto abstract geometric objects that are related as depicted below:

A C

area of square C =

area of square A + B area of square B

DIAGRAM 1

For present purposes, the crucial point about “correspondence” is not thatthe abstract entities exist, but that geometric ideas are not spurious. Since thePythagorean Proposition is provably correct, it somehow “agrees with” reality.(By contrast, endlessly many similar propositions disagree with reality. Con-sider the false claim that the square of the hypotenuse equals twice the sum ofthe other squares.)

Rationalists are struck by several features of cases like these. First, in order toentertain the Pythagorean Proposition, one needs to have the relevant ideas. Ifyou met a creature who didn’t have the idea of a three-sided figure composed ofstraight lines, you couldn’t tell him what the Greeks knew, much less how theycame to know it. In one sense, this is a perfectly general point. If you met some-one who didn’t have any conception of cows, you couldn’t tell him that cowsare brown. But at least you could show him some cows (distinguishing themfrom rocks, horses, bulls, etc.); whereas you can’t show anyone a triangle.6 Soone question is how humans can even entertain thoughts concerning triangles.How do we come to have the relevant ideas?

A recurrent empiricist suggestion is that thinkers somehow abstract ideasfrom a series of sensory experiences, by representing similarities (and ignoringdissimilarities) across those experiences. Whatever the virtues of this proposalas it applies to acquiring the idea of a cow, it is unclear how the suggestionapplies to triangles. How does one abstract the idea of a perfectly straightline from encounters with perceptible objects whose edges are never perfectlystraight? Not only are geometric lines not common features of experience,they are never manifested in experience at all. A possible reply is that thinkerssomehow abstract the ideal of a straight line by observing various nearly straightlines that diverge from the ideal in statistically random ways; in effect, one


Innate ideas 169

“averages” across a range of perceptible cases. But this raises the question ofwhy humans focus on some averages and not others. Why does our idea of aline abstract away from thickness altogether, as opposed to representing someideal or average thickness? More generally, why do human thinkers naturallyform ideas that correspond to certain (averageable) dimensions of perceivedentities, but not others?

Shifting back to language for a moment, this is the point of Chomskyanexamples like (1–4).

(1) John ate an apple.

(2) John ate.

(3) John is too clever to catch a fish.

(4) John is too clever to catch.

Assuming that most speakers of English hear sentences like (1–2) before hearingsentences like (3–4), why don’t they generalize as follows: since (2) meansthat John ate something or other, (4) means that John is too clever to catchsomething or other? Of course, speakers don’t interpret (4) on analogy with (2).They know that (4) means, roughly, that John is too clever for relevant partiesto catch John. One can say that speakers interpret (4) on analogy with someother experienced paradigm(s). But this just pushes the question back. Why dospeakers analogize along some dimensions but not others? The facts suggestthat the explanation will take a rationalist form: the cognitive apparatus broughtto experience has a certain character in virtue of which human beings aredisposed to project from experience to certain characteristic states of linguisticknowledge.

An empiricist account, based on abstraction of ideas from sensory experi-ence, also requires that each thinker who acquires a certain idea has enoughexperience to abstract that idea. Encounters with one or two triangular fig-ures are presumably not a sufficient basis for acquiring an idea that has been“shaped” by experience.7 Moreover, theorists cannot rest content with theclaim that someone could abstract the idea of a triangle (or a verb phrase)given suitable experience. One has to explain why so many thinkers abstractthe same idea across the range of actual experiential situations. If thinkersof varying intelligence regularly form certain ideas, even when their experi-ence is limited, that is a fact to be explained. So as rationalists have alwaysstressed, if humans can acquire certain knowledge quickly and in the absenceof much experience, such knowledge would seem to be a reflection of thecognitive resources humans bring to experience. Plato thus describes hisfamous thought experiment in which an uneducated servant comes to know thePythagorean Proposition with very little experience (given some prompting bySocrates).8



Summarizing the key steps, the servant comes to “see” that doubling the sidesof a square quadruples the area, as suggested by the diagram below:

DIAGRAM 2

He also agrees that a square can have any area. (It is striking that this is soobvious. How would you teach it to someone not initially disposed to agree?)So for example, a square can have an area twice that of the square we startedwith; and doubling the sides of a square with area 2 yields a square with area 8.

2area of each square x = 2

area of the large square = 8

DIAGRAM 3

Socrates then leads the servant to a crucial (but intuitively obvious) lemma:bisecting a square yields two right triangles with equal areas; and so any square,including our square of area 8, can be divided into four right triangles with equalareas.


area of each triangle = 2 2

DIAGRAM 4

If we now extend the figure, by adding a right triangle to create a square of area4, we have an illustration of the Pythagorean Proposition: the square of area 8is the square of the hypotenuse of a right triangle (with area 2), each of whosesides has a square with area 4:


area of the small square = 2 + 2 = 4

22

DIAGRAM 5


Innate ideas 171

And as Meno recognizes, the same result can be achieved (in the same way)for any right triangle with equal sides.9

Faced with the fact that humans can manifest rather esoteric knowledge soquickly, Plato conjectured that humans “recollect” knowledge acquired in somepast (immaterial) life in which ideal objects could be apprehended directly. Butthis is no explanation, absent an account of how immaterial beings apprehendtriangles (and then store this knowledge in a form accessible to material beings).So one may as well – and Chomsky does – treat Plato’s talk of recollectionas a placeholder for unknown cognitive resources from which the knowledgesprings. This makes it clear that Plato’s “experiment” raises more questionsthan it answers: what is the best theoretical characterization of the relevantcognitive resources; how do they give rise to the knowledge in question; andin virtue of what do human beings have these resources? But this is hardly anembarrassment. Figuring out which questions to ask is a crucial starting pointfor inquiry.10

It is also striking that the Pythagorean Proposition, like many mathematicalclaims, is provable from compelling principles that are intuitively more basic.One cannot prove that cows give birth to cows, or that heat rises. One discov-ers, in large part by making repeated observations, that these generalizationsare correct. But not only are measurements of perceptible figures unnecessaryfor geometric knowledge, they are irrelevant. One can’t come to know thePythagorean Proposition, in the way that geometers knew it, by generalizingfrom thousands of carefully drawn and measured triangular figures. A diligentstudent might note that the more carefully she draws and measures, the closershe approaches the ideal. But this is not a proof of the theorem. And prov-able propositions seem to have a distinctive character that is related to the feltnecessity of such propositions.11

If thinkers recognize certain propositions as noncontingent, this suggests thatjudgments concerning such propositions go beyond experience of the mind-independent world. A related observation (Chomsky 1986) is that the samekind of felt necessity attaches to claims like (5–8).

(5) If Mary persuaded John to go to college, John intended to go to college.

(6) If John boiled the soup, the soup boiled.

(7) If John thinks that Bill likes himself, John thinks that Bill likes Bill.

(8) If John is too clever to catch, John is too clever for relevant partiesto catch John.

And speakers of English know that (9) and (10) are not even possible sentencesof English.



(9) Bill thinks that Mary likes each other.

(10) Who did Mary wonder that saw Bill?

These facts about speakers’ capacities to tell that certain strings of words are(not) legitimate English expressions, or that certain English sentences are sureto be true come what may, call for explanation. They invite the hypothesis thathumans (unconsciously) deploy various linguistic ideas that spring from under-lying cognitive resources, with the result that (5–8) are perceived as analyticwhile (9–10) are perceived as illicit.

Cartesian ideas

Knowing a language, however, requires experience in a way that knowledgeof geometry does not. Again, knowing Japanese differs from knowing English,even if the underlying mental states have much in common. Correlatively, know-ing a language requires at least some experience with its vocabulary and certainaspects of its syntax. For example, one has to learn that direct objects typicallyfollow verbs in English, while they typically precede verbs in Japanese; andthat Japanese has overt case marking, as in Latin, while English does not. Thus,while Chomsky regularly speaks of Plato’s Problem, he does not push the anal-ogy to mathematical knowledge too hard. As his favorite seventeenth-centuryrationalists argued, symptoms of innate cognitive resources are not limited toour judgments about abstract objects.

In a famous example, Descartes asks what underlies his various judgmentsconcerning a piece of melting wax. The perceptible qualities of the wax – itsvisual appearance, smell, and so on – change over time. Indeed, Descartesclaims, no perceptible feature of the wax remains constant throughout all pos-sible changes in the wax.12 Descartes concludes that he conceives of the waxas something with (changeable) perceptible qualities, even though the waxdoes not seem to be any bundle of perceptible qualities. The wax seems to besomething that underlies its perceptible qualities. Perhaps this intuitive viewis wrong; though for present purposes, the metaphysics doesn’t matter. Focuson the question of how we come to have the idea of an extended substance:something that occupies space and is a potential bearer of various percepti-ble qualities at various times. What in experience could provide the basis forthe idea that the things we see – like lumps of wax – are things over and abovebundles of sensible qualities?13 Cartesians, seeing no prospects for a plausibleanswer to this question, conclude that even the ideas humans employ whenthinking about concrete perceptible objects are ideas that transcend perceptualexperience.


Innate ideas 173

Descartes applied similar considerations to the idea of a thinking thing;and while one can object to both his metaphysical and psychological conclu-sions, it is easy to see why Chomsky is impressed with the style of argument.Descartes tried to figure out, on the basis of the kinds of judgments we makeabout thinkers (including ourselves), what human beings naturally assume aboutthinkers/persons independent of experience. The goal is to characterize the cog-nitive resources that humans bring to their experience of persons; so Descartestried, as best he could given the theoretical tools available, to isolate our ideaof a person and factor out the contribution of experience. Similarly, Chomskyuses the linguistic judgments of speakers as data for claims about what childrenassume about language, prior to experience.

For various reasons, moral knowledge was also an important topic in theseventeenth and eighteenth centuries. So much of the discussion about innateideas concerned the kinds of moral judgments humans are apt to form. RalphCudworth and Herbert of Cherbury, mentioned above, held (roughly) thathuman nature makes a certain range of moral ideas naturally available to any-one who grows up in a normal human environment. And once a thinker startsto form judgments involving these ideas, particular judgments – say, “killingpeople is wrong” – will be especially compelling. While early formulationsof this view were often incautious, and open to easy criticism by empiricists,Chomsky endorses the underlying picture.14 Human minds develop in accor-dance with human nature, projecting ideas in response to initial experience;and then humans use those ideas, as best they can, to represent a world that haswhatever character it has. In certain domains, like geometry and language andmorals, these ideas manifest themselves as a special kind of knowledge whosesource is “from within.”

One might think there is a crucial difference between the “linguistic ideas”that seem to spring from human nature and ideas of perceptible objects. For thelatter are at least “directed” towards the language-independent world, even ifhuman judgments fail to represent that world accurately. And one might thinkthat ideas of external objects – ideas that are (or at least are intended to be) aboutthings independent of those ideas – cannot be innately constrained in the waythat the idea of a verb phrase is constrained. But stated this baldly, the objectionjust restates the denial of rationalism. On Chomsky’s own view, there are nolinguistic expressions independent of language users; nothing “outside us” isrelevant to whether or not our natural linguistic judgments are right or wrong.Correspondingly, an ordinary speaker’s idea of a verb phrase is not about someaspect of the external world to which the speaker is attuned. On the other hand,an ordinary person’s idea of (say) wax is about – or at least is sometimes used tothink about – some aspect of the mind-independent world. But the fact that wethink about things outside us hardly shows that our ideas owe their character to



those external things. And the fact that we occasionally “tune in” to features ofthe mind-independent world, in ways that let us form reliable judgments, hardlyshows that our ideas are not predominantly shaped by the cognitive systemsthat gave rise to them.

From this perspective, we are fortunate if we can use our natural ideas toform thoughts that truly reflect mind-independent reality. So one should notassume that humans have a natural capacity to form reliable judgments aboutevery aspect of the environment that attracts their attention. Thus, rational-ists typically stress that the very capacities that make knowledge possible alsoimpose constraints on what one can know by deploying those capacities. Forexample, the language faculty imposes constraints on which languages humanscan naturally acquire. This leaves open the possibility of “decoding” a non-human language – i.e. a language that does not conform to the principlesof universal grammar – by using cognitive systems other than the languagefaculty; but we could not understand such a language in the way we can under-stand human languages. Similarly, when acting as scientists, humans may beable to construct ideas that lie outside the space of ideas naturally availableto them; although Chomsky (1975, 1980, 2002b) suggests, in good rationalistfashion, that whatever cognitive systems underlie our capacity for doing sci-ence will also be governed by their own internal principles (see also McGinn1994a).

That said, Chomsky also thinks that by pursuing the kind of scientific researchthat linguists pursue, we can begin to understand aspects of how human beingsrepresent and reason about the world. So we conclude this chapter with a briefdiscussion of two phenomena that seem to straddle syntax and semantics andlogic. For such phenomena may provide suggestions about how to (slowly)push beyond the study of linguistic forms, and into the study of why linguisticexpressions have the meanings they do (as opposed to other logically coherentbut “non-human” meanings), while remaining focused on how human ideas areconstrained by our biological endowment.

Beyond mere grammar

As we have emphasized, Chomsky contends that the cognitive apparatus humanbeings bring to language learning has a certain character, given which humansacquire certain characteristic states of linguistic knowledge that go well beyondtheir experience. To illustrate this point about the language faculty, Chomskyfrequently notes that as soon as children are able to combine the words brownand house to form the expression brown house, the composed phrase has seman-tic properties beyond those attributable to the meanings of the individual words.In particular, children use That is a brown house to say that the exterior ofthe house in question is brown; and unless interiors are explicitly mentioned,


Innate ideas 175

children will hear That is a brown house as a claim about the exterior of thehouse. (See chapter 10.)

The kind of innate knowledge that Chomsky envisions also includes knowl-edge of linguistic properties that have been investigated by researchers in thefield of formal semantics. Chomsky views formal semantics as a theory abouta component of the language faculty that determines certain features of struc-tures generated by the computational system – i.e. the syntax. These structuresinclude expressions of various categories. For example, the category of deter-miner covers a large class of (intuitively quantificational) expressions like some,all, no, every, most, at least three, more than three, but less than ten, and soforth.

There seem to be semantic universals that constrain the contribution of deter-miners to the meanings of expressions that contain them. One proposed uni-versal property of determiner meanings, known as conservativity (Barwise &Cooper 1981; for introductory discussion, see Larson & Segal 1995; Chierchia& McConnell-Ginet 2000), is illustrated in (11).

(11) If all cows are brown cows, then all cows are brown.

Speakers of English can know that (11) is sure to be correct, regardlessof the situation. Moreover, (11) remains obviously correct if we replace allwith other determiners like some, no, and so on. This suggests the followinggeneralization:

(12) If [(DET cows) (are brown cows)], then [(DET cows) (are brown)]

In (12), “DET” can be replaced by any natural language determiner to producean obvious truth. Some examples include those in (13).

(13) a. If most cows are brown cows, then most cows are brown.b. If all but two cows are brown cows, then all but two cows are brown.c. If more than three cows are brown cows, then more than three

cows are brown.

This turns out to be an important fact. To see why, it will help to note that onecan specify the meaning of a determiner using (something like) set-theoreticrelations. One can think of determiners as names for relations that can holdbetween sets. For simplicity, let’s continue to focus on sentences of the formDET NP VP, where the determiner is followed by a noun (or noun phrase) likecows and a verb (or verb phrase) like are brown cows; the NP and VP are,respectively, the first and second arguments of the determiner. A sentence ofthe form DET NP VP is true if and only if the set associated with the first (NP)argument bears the relation named by DET to the set associated with the second(VP) argument. For example, the sentence All cows are brown is true if and



only if the set of cows is a subset of the set of brown things. So one can thinkof the determiner all as a label for the subset relation. Likewise, All cows arebrown cows is true if and only if the set of cows is a subset of the set of browncows. So (11) expresses the following set-theoretic truism: if the set of cows isa subset of the set of brown things, then the set of cows is a subset of the set ofbrown cows.

The sentence Some cows are brown is true if and only if the set of cowsintersects with the set of brown things. So one can think of the determiner someas a label for the relation of intersection. Likewise, Some cows are brown cowsis true if and only if the set of cows intersects with the set of brown cows;and if the set of cows intersects with the set of brown things, then trivially,the set of cows intersects with the set of brown cows. Similar remarks applyto all natural language determiners. The generalization indicated in (12) isquite robust. But natural languages also contain quantificational expressionsthat would be counterexamples to (12) if these words were determiners. Forexample, inserting the word only into (12) would yield (14), which is not atruism (as indicated by ‘#’).

(14) # If only cows are brown cows, then only cows are brown.

Suppose there is a brown horse in the conversational context. Then while (11)and (13) remain obviously correct, (14) is false. So to maintain the generalizationin (12), one must say that only is not a determiner. But this is independentlyplausible, since only (which is arguably an adverb) can appear in a much widerrange of syntactic positions than genuine determiners. Consider, for example,He only likes brown cows and Bessie is the only brown cow.

The availability of words like only in natural languages shows that logic alonedoes not explain the generalization about the conservativity of determiner mean-ings. Rather, it seems that natural language is intolerant of determiner meaningsthat would violate (12), though it tolerates such meanings for other kinds oflinguistic expressions. So even if actual instances of (12) report logical truths,logic alone does not exclude the possibility of quantificational expressions thatwould be counterexamples to (12). The relevant generalization evidently holdsbecause of some fact about speakers of human languages, whose mental gram-mars are subject to a substantive constraint that prohibits certain determinermeanings; see Pietroski (2005) for a proposal.15

If this constraint turns out to be a linguistic universal, then it is a likely candi-date for innate specification. For what evidence (that children are sensitive to)would allow speakers to infer that other speakers use only conservative deter-miners? The nativist proposal would be that the conservativity of determinermeanings is a reflection of features shared by all natural languages in virtueof human biology. These “core” features of natural languages constitute whatChomsky calls “Universal Grammar” – the initial state of the language learner. A


Innate ideas 177

consequence of Universal Grammar is that sentential structures are interpretedin certain ways. If this is correct, then theorists can try to characterize the notionof “natural” consequence that is a by-product of how the human language facultyis organized (see Ludlow 2002). And it seems that the “interpretive” compo-nent of Universal Grammar does indeed impose constraints on the meanings ofso-called logical words. A child exposed to a particular spoken language hasto figure out which sounds in that language are associated with which possiblemeanings for quantificational expressions like all, some, and no. But the nativistexpects a child to know how such words contribute to the truth conditions of sen-tences (in any natural language) as soon as these words have entered the child’slexicon. For example, the nativist will expect the child to manifest knowledgethat all determiner meanings are conservative, as soon as it is possible to testfor such knowledge.16 (We return to words for propositional operators – and,or, and not – presently.)

Other semantic properties of determiners are also viable candidates for innatespecification. Although all determiners are conservative, they can be furtherpartitioned into semantic classes that correspond to certain entailment relationsamong sentences. Consider the impeccable inferences in (15):

(15) NP: If no cow ate a vegetable, then no brown cow ate a vegetable.VP: If no cow ate a vegetable, then no cow ate a green vegetable.

Given the determiner no, one can replace a general term like cow with a morespecific expression like brown cow in the first (NP) argument. One can alsoreplace a general term like vegetable with a more specific expression like greenvegetable in the second (VP) argument. To capture this fact, semanticists saythat no is “downward entailing” in both of its arguments. This is related to thefact that negation generally licenses inferences from claims about sets of thingsto claims about subsets. For example, if John did not buy a car, it follows thathe did not buy a red car; whereas if John bought a car, it does not follow thathe bought a red car.

But not all determiners have this feature, as indicated in (16–17):

(16) NP: If every cow ate a vegetable,then every brown cow ate a vegetable.

#VP: If every cow ate a vegetable,then every cow ate a green vegetable.

(17) #NP: If some cow ate a vegetable,then some brown cow ate a vegetable.

#VP: If some cow ate a vegetable,then some cow ate a green vegetable.



The universal quantifier every is downward-entailing on its first argument, butnot on its second argument. The existential quantifier some is not downwardentailing on either of its arguments.

The semantic property of downward entailment has important linguistic con-sequences. For it is connected to several seemingly unrelated linguistic phenom-ena, including the licensing of so-called negative polarity items (such as any,ever, and at all) and restrictions on the interpretation of or. The sentences in(18–20) illustrate that the negative polarity item any is licensed by downward-entailing argument positions. For example, every is downward entailing on itsfirst (NP) but not its second (VP) argument. And as shown in (19), any canappear in the complex NP cow that ate any vegetable; but if any appears in thecomplex VP ate any vegetable, it sounds odd (given the determiner every), asindicated by the asterisk. By contrast, no is downward entailing in both of itsarguments, and any can appear in either argument; while some is downwardentailing in neither of its arguments, and any cannot appear in either argument.

(18) NP: No cow that ate any vegetable became ill.VP: No cow ate any vegetable.

(19) NP: Every cow that ate any vegetable became ill.VP: ∗Every cow ate any vegetable.

(20) NP: ∗Some cow that ate any vegetable became ill.VP: ∗Some cow ate any vegetable.

Similar remarks apply to the complex NP cow that ever ate vegetables and thecomplex VP ever ate vegetables. The generalization seems to be that negativepolarity items are licensed in downward-entailing environments.

Another striking fact is that if a determiner is downward entailing on oneof its arguments, the disjunction operator has a “conjunctive” implication inthat argument position. This is illustrated in (21), where or has a conjunctiveimplication when it appears in the first argument of every, but not when itappears in the second argument.

(21) NP: If every cow that ate broccoli or asparagus became ill, then(i) every cow that ate broccoli became ill and(ii) every cow that ate asparagus became ill.

#VP: If every cow ate broccoli or asparagus, thenevery cow ate broccoli andevery cow ate asparagus.

The conjunctive implication is present for both arguments of no and neitherargument of some. For example, if no cow ate broccoli or asparagus, it followsthat: no cow ate broccoli, and no cow ate asparagus. But if some cow ate broccoli


Innate ideas 179

or asparagus, it does not follow that some cow ate broccoli; nor does it followthat some cow ate asparagus.

One can think of DeMorgan’s law for negation, stated in (22), as a specialcase of a general relation between disjunction and conjunction in downwardentailing linguistic environments, as stated in (23).

(22) not(A or B) ➔ not(A) and not(B)

(23) For any downward-entailing operator O: O(A or B) ➔ O(A) and O(B)

The generalization in (23) extends to all downward-entailing expressions. Butsince every is downward entailing only in its first argument, the conjunc-tive implication of disjunction arises only in its first argument. Theorists canthus capture the fact that every falls between no and some with regard to theDeMorgan phenomenon.17

In thinking about such facts, it is important to distinguish logical truths – aboutthe interrelations of logical notions like conjunction, (inclusive) disjunction, andnegation – from facts about the (natural) meanings of linguistic expressions.The logical truth reported with (22) does not, by itself, tell us anything about themeanings of sentences involving the natural language expressions not, and, andor. In this regard, note that DeMorgan’s law would not be germane to sentencesof the form A or B if the word or had an “exclusive” interpretation accordingto which A or B is false if A and B are both true. This suggests that certainexpressions have the property of being downward entailing because speakersof human languages are subject to a substantive constraint involving downwardentailment.

This constraint is relevant to at least all of the following: the basic (inclusive)meaning of disjunction words in natural languages, the licensing of negativepolarity items, and prohibitions on the imposition of scalar implicatures. Theconstraint under consideration is another candidate for innate specification. Ifso, language learners are expected to approach the task of grammar formationequipped with this aspect of logical reasoning. Whatever “errors” of logicalinference learners make, we do not expect them to violate DeMorgan’s laws,or to produce negative polarity items in the wrong linguistic environments.Recent experimental evidence from studies of child language lends credence tonativist hypotheses. As soon as children can be tested, around the age of three,they obey the licensing conditions on negative polarity items, they computethe conjunctive interpretation of disjunction in downward entailing linguisticenvironments, and they evince knowledge that determiner meanings are con-servative (Chierchia et al. 2001).

This is the kind of inquiry suggested by Chomsky’s conception of innateideas. One tries to learn about the character of human thought by lookingfor generalizations that are neither logical truths nor plausible candidates for



hypotheses that thinkers have empirically confirmed on the basis of data avail-able to them. If young children, with different backgrounds, all respect a (non-logical) generalization G that trained linguists have only recently noticed, thissuggests that G is a reflection of human nature. This is then a starting pointfor further inquiry into how (and why) human nature gives rise to such gen-eralizations. Chomsky – following Plato, Descartes, and others – thus offers amethodology for how to formulate and occasionally answer substantive ques-tions about human thought.


9 Mind, language, and the limits of inquiry

Akeel Bilgrami and Carol Rovane

This chapter explores a very general philosophical and methodological theme inNoam Chomsky’s work – the scope and limit of scientific inquiry in the study ofmind and language. It is a conspicuous fact about Chomsky that accompanyingthe vast and driving intellectual ambition of his program in what he conceivesas the science of linguistics is a notable and explicit modesty about the extentto which he thinks he has given, indeed the extent to which one can give,scientific answers to fundamental questions. This modesty in terms of breadthof coverage is in a sense the other side of, and therefore indispensable to, thedepth of what he has achieved in the area he has covered.

In his work, he seems to offer at least two different sorts of reasons for us to bemade modest about ourselves as inquirers. First there is a modesty implicit in hisguardedness about claiming for semantics what some other philosophers haveclaimed for it, and what he himself has claimed only for syntax understood ina broad sense viz., that there is in some interesting sense an explanatory theoryto be offered which can be incorporated into the science of linguistics. Second,there are reasons for modesty having to do with the fact that either because ofour conceptual limitations or because of faulty formulations of questions, weare in no position to give serious and detailed answers to them. The next twosections will take up each of these in turn.

Is referential semantics possible?

For Chomsky, scientific inquiry into language and into the human mind ispossible if it can assume that what is being studied are the “inner mechanisms”which enter into the study of thought and expressions and behavior generally.As he says:

The approach is “mentalistic” but in what should be an uncontroversial sense. It isconcerned with “mental aspects of the world” which stand alongside its mechanical,chemical, optical, and other aspects. It undertakes to study a real object in the naturalworld – the brain, its states, and functions – and thus to move the study of the mind [andlanguage] towards eventual integration with biology and the natural sciences. (Chomsky2000a: 6)

181


182 Akeel Bilgrami and Carol Rovane

Though eventual integration with biology is the goal, it is a distant goal. In theinterim scientists work with the data and the theoretical resources available tothem, at a level of description and explanation which it allows them. They havethe scientific goals of describing and explaining the language faculty whichis present in the entire species as a biological endowment, but at a level ofdescription and explanation which in the interim is bound to be a cognitiveand computational level, with the properties of internality, universality, innate-ness, domain-specificity, among others, all of which Chomsky’s own successivetheories of grammar over the last few decades have exemplified.

This deep commitment to internalism is presented as being of a piece withwhat Chomsky says is the naturalistic intractability of semantics as standardlyconceived, which relies heavily on reference and more generally on the relationsour words and concepts bear to objects, properties, and states of affairs in theexternal world. Two main reasons emerge for this skepticism from a numberof interesting remarks over many essays. First, we have extremely rich anddiverse conceptions of the things our words refer to, and that infects referenceitself, making it a highly mediated and contextual notion. This thwarts scientificgeneralizations about reference from ranging over all speakers of a naturallanguage and even perhaps over a single speaker at different times. And second,there is no reference without speakers intending to refer, and intentionality ingeneral is not a fit subject for naturalistic treatment. Let’s look at each of thesein turn.

In stressing agents’ rich and diverse conceptions of the things they refer to,Chomsky resists a normative as well as a social understanding of the notionof reference. He repeatedly rejects the intuitions urged by both the proponentsof twin-earth thought experiments as well as socialized variants of it such asBurge’s highly fortified example about his protagonist’s arthritis. And he con-cludes, rightly in our view, that there is no theoretical compulsion to insistthat the term water used on twin-earth and earth must always have differentmeanings and reference (for example, even for speakers here and there, whoknow no chemistry), nor to insist that the term arthritis on the lips of Burge’smedically ignorant protagonist must mean and refer to what the doctor’s termin his society means and refers, rather than to a wider class of ailments.1 Socialand other external relations do not force a uniform norm of meaning and ref-erence of a term on all speakers of a language, such that all departures fromit necessarily amount to mistakes. For some departures, instead of thinking ofthem as violations of a norm, we can think of them as individual (“idiolectical”)meanings and references, tied to local contexts of use.

There might be two different referentialist responses to this appeal to thediverse conceptions of things to which we refer with our terms.

The first would be to say that despite the diverse conceptions that speakershave, they all intend to use a term as others do; they all intend their use of a


Mind, language, and the limits of inquiry 183

name like Hesperus, or a natural kind term like water, to refer to what others,especially the experts, in the community refer to. Or (a somewhat differentaccount) they may intend to refer to that thing which was named by the originarybaptismal reference-fixing event, or instances of that substance which have thesame scientific nature as the substance picked out in the originary, reference-fixing event. These intentions give uniformity to the reference of these termsfor all speakers who use them, so no dreaded contextuality arises from thediversity of conceptions speakers might have of the things they refer to. Overmany essays (some are found in his [2000a]), Chomsky addresses all theseaccounts and has trenchant things to say against them. First of all, he pointsout, the data leave underdetermined whether one should think of reference ashaving this uniformity or think of it instead as being much more contextual andindividual. Certainly data about deference among speakers towards expertsin the community do not necessarily point to a socially constituted notionof reference because they can be handled quite easily within the idiolecticalapproach to reference by simply pointing out that the reference of an individual’sterm changes once one learns from experts and defers to them. And then, hepoints out, quite apart from data not forcing the issue, none of the theoretical orphilosophical motivations philosophers have had for stressing such a uniformand decontextualized notion of reference is compelling either. He patientlyaddresses such motivations (e.g. that only such a notion will account for theory-change as being distinct from meaning-change, and for how one may learn aboutthe world – and not just about what is intended by the speaker – from others’usage of terms) and shows that these things are all easily accounted for withinan individualist approach.

But even putting those criticisms aside, his point remains that these accountsachieve their uniformity and transcend particular contexts only by relying onthe intentions of speakers and – intentionality being what it is – that puts themoutside what is naturalistically tractable in a theory. In fact, both Kripke andPutnam who favor this form of referential semantics are careful to make noclaim to a theory of reference, leave alone a naturalistic and scientific accountof it.

But there is a second referentialist response which, realizing the naturalism-thwarting element of the first’s appeal to intentions, does not appeal to intentionsin its account of meaning and reference. This is the view, owing first to Dretskeand much refined and developed by Fodor, which ties reference to causal covari-ances between mental tokens of a type in the language of thought and objectsor properties in the world (Fodor 1975, 1998; Dretske 1981; Frege 1892). Onthis view, the rich and diverse conceptions of things that speakers may haveof the objects referred to are irrelevant because the causal relations posited areuncontaminated by such mediating conceptions. And so the sorts of intentionsappealed to by the first response in order to finesse these conceptions of things



are unnecessary. There is no question, in any case, of appealing to intentions torefer since we do not and cannot have intentions towards terms in the languageof thought, we can only have them towards words we vocalize. Since neitherintentions nor conceptions of things play any role, these relations between a term(concept) and an object or property in the world may be the basis of universallaws which hold for all speakers who possess the concept and who stand incausal relations with the object or property in question. In fact Fodor (1990)sometimes himself describes the aspirations of such a naturalistic semantics inNewtonian terms. In a sense, this second challenge to Chomsky is the moreinteresting one because it accepts one half of his overall view (the naturalismand the scientific aspirations for linguistics) and resists the other half (the inter-nalism, or the claim that it is only internalistically described phenomena whichare scientifically tractable). This referentialist response holds that reference isscientifically tractable, and therefore there is a respectable naturalistic versionof semantics, as well as syntax.

Despite the fact that Chomsky does not explicitly say so, we suspect thathe would be unimpressed by this response which, while it does in a very gen-eral way allow for naturalism (purely causal covariances), it does not offerany specific suggestions for naturalistic inquiry, no specific research programs,no specific hypotheses, no design for specific experiments to test hypotheses.There is only an assertion that the subject of reference has very austere causalcovariances underlying it which involve no conceptions, intentions, etc., andthat it provides no obstacles to naturalism and the search for general decontextu-alized scientific laws in the study of semantics. Chomsky’s successive theoriesof universal grammar, all of which restrict themselves to syntax broadly under-stood, are rich and detailed. Fodor’s naturalistic referential semantics is, bycontrast, little more than a suggestive idea. It seems very much the suggestionof a philosopher straining to make claims for reference that lie within science,but with no real sense of what science must actually then do in this area ofstudy.

But even putting this important qualm aside, there is another worry muchmore on the surface of what Chomsky does explicitly say in the many passageswhere he speaks of our ordinary concept of reference: he is bound to ask ofFodor’s naturalistic version of “reference,” why is this an account of reference?Why is it not to be seen as giving up the idea of reference for causal covariances?He may not have anything against such a naturalization (Chomsky does not ingeneral feel any qualms about changing the subject from common sense toscience), so long as it is not claimed that it is reference that we are still talkingabout. There must be some common features, some shared structure, betweenreference as ordinarily understood and reference as naturalistically understoodin these terms, which makes it clear that the notion is indeed preserved morethan nominally.



What makes Fodor particularly interesting as an interlocutor is that he explic-itly argues that something deep is preserved. His notion of meaning and (inten-tional) content is based exclusively upon the notion of denotation or reference.(As he says at the beginning of his [1990], “The older I get the more convincedI am that there is no more to meaning and content than denotation.”) And, inturn, meaning and content are what go into the explanations involved in what hecalls “granny psychology,” the psychology which cites content-bearing states inthe explanation of intentionally described behavior. One’s intentional contents,contents specified in that-clauses (the belief that water quenches thirst, say) areindividuated strictly by the referents of the component terms, such as the termwater. Reference, even after it is naturalistically characterized in terms of thecausal covariances that hold between our mental tokens of water and instancesof a substance with a certain chemical composition, continues to contribute tocontents of intentional states such as the one just cited which (in intentionalpsychology) explain actions of ours in the world, such as drinking water whenwe are thirsty, etc. And it is part of his claim that this psychology, the psychol-ogy whose states are expressed and understood by the grannies of the world, isnot to be “eliminated” at all for another psychology, which makes no mentionof intentional content. Rather, granny psychology approximates the truth (ortruths) eventually captured in full naturalistic dress when one sees through itschief notions (content, meaning, reference) to what underlies them – the causalcovariances.

We have here the real target of Chomsky’s skepticism. What he is rejectingis the idea that when we come up with these universal laws based on causalcovariances – granting for the moment that these are the deliverances of aninteresting scientific research project, which is doubtful – we have come upwith something that is in any way interestingly continuous with intentionalpsychology as understood in common sense.

Chomsky says many things that make it clear that he would be skeptical. Hereare two related arguments that support his skepticism. He does not formulatethese arguments in just these terms, but it seems to us that they drive his doubts.

First argument. We are considering a naturalism about reference which alsoclaims that reference plays a vital and exclusive role in the attribution of inten-tional content and generally in intentional psychology. Now, any view of ref-erence (of our terms or of the concepts which those terms express) should becompatible with the following constraint on the commonsense attribution ofintentional content to a subject: if a subject believes something with an inten-tional content or expresses that belief with that intentional content by utteringa sentence, and that belief (or assertion) is merely false, i.e. if the speaker ismerely misinformed about something in the world, it should still follow thathe is quite rational in having that belief with that content (or in making thatassertion). In other words, merely being misinformed does not bring with it



irrationality or incoherence. For example, suppose someone is misinformedabout the chemical composition of water and he says, “Water is not H2O” (orhas a belief with the content that water is not H2O). Now if the reference of theterm (or concept) water is given by the causal covariance between his relevantmental tokens and instances of H2O, then it strictly follows that he is think-ing something inconsistent. But this man is merely misinformed in saying orbelieving what he does. He is not irrational and logically incoherent. Thus thenaturalistic view of reference violates a basic constraint on our commonsenseunderstanding of reference and its role in intentional psychology. A naturalisticpsychology based on such an understanding of reference which violates thisconstraint therefore fails to be continuous with ordinary intentional psychology,such as it is.

It should be apparent that this argument echoes, indeed that it more or lessis, Frege’s argument for sense. Chomsky appeals explicitly to Frege and usesthe term “perspectives” instead of “senses.”2 Frege and Fregeans go on to spoilthis famous argument and this important constraint on which it is based bydemanding all sorts of further things of the notion of sense: viz., that sensesare abstract objects to which our thinking is related, that to be this they must beexpressed in a shared language, etc. – all claims of which Chomsky is critical.We come back to that in a moment.

Second argument. To repeat, we are considering a naturalism about referencewhich also claims that reference plays a vital and exclusive contributing role inthe attribution of intentional content and generally in intentional psychology.Now, any view of reference (of our terms or of the concepts which those termsexpress) should be compatible with the following constraint on the attribution ofintentional content to a subject: if a subject believes or desires something (say,believes that drinking water will quench his thirst or desires that he drink thewater in front of him), and there are no familiar forms of psychological obstaclessuch as self-deception or other less interesting psychological obstacles such asthat it is simply too submerged in his thinking, then he knows what he believes ordesires. Of course, self-knowledge does not hold ubiquitously of our beliefs anddesires precisely because we have many beliefs and desires which we repressor which are too submerged in our psychologies, etc. But we can assume thatif these psychological obstacles and censors are not present, then awareness orself-knowledge of the intentional states would be present. Its presence could notbe denied by anything but such internal psychological obstacles. It could not bedenied by philosophical fiat, it could not be denied because Fodor has proposed acertain theory of reference. Let’s take an example. Suppose someone is ignorantof chemistry, in particular of the chemical composition of water. And supposehe says (or believes) “Water quenches thirst.” If the term or concept water inthat assertion or belief has the reference it has because of the causal covariancewhich holds between the mental tokens of the relevant mental type and instances



of H2O, then this subject believes (says) something of which he is quite unaware,i.e. he believes that a substance with a certain chemical composition quenchesthirst. He could not possibly be aware of what he believes since he knowsno chemistry. But not knowing chemistry is not a psychological obstacle ofany kind. It is just ignorance about the world. On this view of intentionalpsychology, this subject, in order to gain self-knowledge of what he himselfthinks, would have not to overcome repressions, self-deceptions, and otherpsychological obstacles, he would rather have to learn more chemistry, learnmore about the chemical composition of substances in his external environment.Thus, the naturalistic view of reference violates another constraint about ourcommonsense understanding of reference and its role in intentional psychology.Again, in violating this constraint, a naturalistic psychology, based on such anunderstanding of reference, therefore, fails to be continuous with intentionalpsychology, such as it is.

Both these arguments are implied by Chomsky’s attitude towards referenceas it figures in our commonsense understanding.

The first, Fregean argument requires that the notion of reference be embeddedin the context of various conceptions of an object to which the speaker intends torefer, just as Chomsky has all along explicitly insisted. The conceptions are notseparable from the object to which the speaker intends to refer. It’s the object,under those descriptions or conceptions, to which the speaker refers. Chomskytends to assimilate any view which denies this embedding as the “myth of thelogically proper name.” One function of embedding reference in conceptions isto make rational sense of the speaker who is merely misinformed (in the caseswe are discussing, misinformed about various a posteriori identities, e.g. theidentity of Hesperus with Phosphorus, or of water with H2O, etc.). Without thisembedding, the constraint on intentional psychology that requires that a notionof reference keep continuity with that psychology is violated.

The second argument is only implied by some things that Chomsky says. Inthe first argument, an agent’s conceptions of things, or what Frege called senses,were seen to be essential to understanding reference, and to the intentional psy-chology of agents to which the reference of terms (or concepts) contributes. It isessential, as we said, because without it, someone who was merely misinformedabout identities would be viewed as being self-contradictory. What philosopherscall “Frege puzzles” about identity are based on this. Someone who does notknow that Hesperus is Phosphorus may think that Hesperus is bright and Phos-phorus not bright, and we know that the person is not contradicting himself.So we have to introduce senses or conceptions of things (“perspectives”) asindividuating his concepts and terms, rather than reference, in order to makehim come out as rational. But, how would we have to view him if we thoughthe was contradicting himself? The idea would have to be that since both termsreferred to the same planet, and since (as Fodor’s naturalism insists) reference,



not senses and conceptions, individuates terms and concepts (in this case, sin-gular terms and singular concepts, but as we saw the point applies equally towater), he must have two contradictory thoughts. He would not of course knowthat he was contradicting himself. It’s not as if he knows that Hesperus andPhosphorus are the same object and he is perversely saying knowingly contra-dictory things about them. Rather, he would be unaware that he was thinkingand talking about the same planet, but he would be talking about them, andthat is why he would be contradicting himself. Self-contradiction in an agentcan be made tolerable only if it is accompanied by such lack of knowledge ofhis self-contradictory thoughts. Thus if senses are left out of the individuationof concepts and the contents of an agent’s thoughts, and if individuation ofconcepts was seen as a matter of reference, the ensuing self-contradiction in histhinking would be tolerable only if it is explained by saying that his ignoranceof astronomy would amount to an ignorance of his own thoughts, his ownintentional (in this case, self-contradictory) psychology. Senses or “perspec-tives” (unlike reference) therefore make sure not only what the first argumentdemands of them, viz., that people merely misinformed (about identities, inthis case) do not come out as having contradictory thoughts, but they also makesure that those thoughts are self-known to the agent. This latter task of sensesis what the second argument demands as a constraint on thoughts.3

What is it about senses which ensures that they will carry out this secondtask, that they will see to it that our intentional psychology, i.e. our intentionalstates, are self-known to us (unless, of course, there are psychological obstaclesto it)? As we just saw, to make things tolerable, one is forced to say that onefails to know what one thinks if what one thinks, or elements in one’s thinking,such as one’s concepts, are individuated by objects (such as, in these examples,planets or cities to which our concepts refer). To put it in terms of language,one can fail to know what one is saying if the meanings of one’s terms arespecified in terms of objects. So if senses are to avoid the problem of leadingto lack of self-knowledge of one’s thoughts, they must precisely not be like thesorts of things which are the source of the problem, which can lead to lack ofself-knowledge. To put it in a word, senses cannot be like planets and cities,they cannot themselves be objects about whose identity we can be misinformed,thinking for instance that there are two senses when there is only one, as wemight do with a planet or a city. If they were objects, we would not be able tosee them as solutions to Frege’s puzzles about identity. They would be subjectto similar puzzles themselves. Thus there are no such things or entities as sensesor thoughts to which we are related in our thinking in the way that Fodor andother referentialists think we are related to objects such as planets and waterin the world. However, Frege, unfortunately, thought that senses are objects,abstract ones. But senses can only do the job they are asked to do by him inthe first argument, they can only do the job that Frege himself wanted them



to do (to solve Frege puzzles about identity), if they also do the job they areasked to do in the second argument, which is to make sure that they and theconcepts and thoughts they individuate are self-known to the agent. And theycan only do this latter job if they are not themselves objects. For if they wereobjects one might be confused into wondering whether some sense of ours wasthe same as another just as one might wonder if Hesperus is Phosphorus, orwater is H2O?4

It is this insight – that thoughts are not objects – which Chomsky explicitlyarticulates against Frege, and in doing so he implies the force of the secondargument given above. He cites much earlier thinkers in support of the insight,saying:

The basic assumption that there is a common store of thoughts surely can be denied;in fact, it had been plausibly denied a century earlier by critics of the theory of ideaswho argued that it is a mistake to interpret the expression “John has a thought” (desire,intention, etc.) on the analogy of “John has a diamond.” In the former case, the ency-clopedist du Marsais and later Thomas Reid argued, the expression means only “Johnthinks” (desires, etc.) and provides no ground for positing “thoughts” to which Johnstands in relation. (Chomsky 1966, 2002a, 1993b: 18)

The insight can now be generalized to make the point that is needed againstthe project of naturalizing intentionality by individuating thoughts and conceptsin terms of the external objects and substances with which we stand in causalrelations. Thoughts are not objects, as Chomsky following earlier eighteenth-century critics is pointing out. We have just seen that there are no internalor abstract objects such as senses. So Frege’s insight of our first argumentabout the need for senses, conceptions of things, perspectives on things, etc.(which Chomsky endorses) would be undermined if we thought that senses werethemselves objects. The second argument ensures that it is not undermined inthis way, by posing a constraint which cannot be met if we take senses to beobjects. But the claim is more general in its significance than that. Our conceptsand thoughts are not individuated in terms of internal objects, but equally theyare not individuated in terms of external objects either. What the referentialsemanticist offers, when he tries to finesse an agent’s conceptions of things,is precisely this externalist individuation. He tries to make the contents of ourthoughts depend on nothing but the external objects with which the conceptswhich compose them stand in causal relations. In saying this he falls afoul ofthe two constraints which we identified in the two arguments above, whichdefine our ordinary commonsense understanding of meaning and reference andintentional psychology. Falling afoul of them makes clear that the continuitywith granny psychology that Fodor himself seeks would go missing.

And Chomsky’s interesting point here is that at a general enough level ofdescription of the mistake, both Frege and Fodor are making the same mistake,



they are both in their different ways individuating thoughts in terms of objects.That they are doing so differently, in terms of internal5 and external objectsrespectively, should not distract from the fact that they make the same mistakeat the more general level.

All this is related directly to Chomsky’s stance on the subject of reference.Frege, in insightfully exposing the flaws in the idea that concepts are individ-uated in terms of the external objects posited by the referentialist, introducesthe importance of the idea of an agent’s conceptions of things, but he does notrest with that insight; he goes on to spoil it by viewing these as internal andabstract objects. And on the other side, the naturalistic referential semanticist,also insightfully acknowledges that conceptions of things would not be the sortsof things that could be naturalistically treatable (Fodor is explicit about this),but then does not rest with that insight; he goes on to spoil it by individuatingthoughts in terms of external objects with which we stand in causal relations,and which he thinks confer naturalism upon reference. Thus the point made byChomsky (and Reid and du Marsais) can be generalized to say that thoughts arenot to be individuated in terms of objects at all, external or internal, and oncewe do so, we can rest with the two insights that were respectively observed byboth Frege and the referential naturalist, and then spoilt by them, when theywould not rest there. We have already italicized them above. They are (1) thereis no understanding reference to things without there being conceptions thatparticular speakers have of the things to which they intend to refer, and (2)conceptions of things are not naturalistically treatable. These, as we have seen,are the very insights that Chomsky has all along insisted on in thinking aboutreferential semantics.

Having argued that reference to things is not the sort of thing that comesunaccompanied by the intentions and beliefs (conceptions of things) of speakers,he argues that reference must therefore really be understood as part of theuse of language. It is not part of the description of the language organ orfaculty, of the mechanisms and internal cognitive system that enable the use oflanguage. It is rather part of a description of what is enabled, which goes muchbeyond a description of the enabling apparatus, involving such things, as wesaid, as a person’s intentions and his richly conceived understanding of whatthe objects around him are, none of which can be the object of “theoreticalunderstanding” and “naturalistic inquiry,” but is rather illuminated by widerforms of understanding which it would be just confusing and conflating to call“theoretical” or “scientific” or “naturalistic.” Since the use of language hastraditionally been seen to fall within pragmatics, Chomsky boldly proposesthe revisionary classification of placing reference not in semantics at all, butin pragmatics. He says, “It is possible that natural language has only syntaxand pragmatics”; and then he adds, quoting an earlier work by himself whichhe says was influenced by Wittgenstein and Austin, “it has a ‘semantics’ only



in the sense of the study of how this instrument whose formal structure andpotentialities of expression are the subject of syntactic investigation, is actuallyput to use . . .” (2000a: 132) This redrawing of the traditional boundaries ofthe trio of syntax, semantics, and pragmatics is by no means arbitrary. Theentire earlier discussion of the nature of reference provides the methodologicalmotivation. Given the fact that the central notion of semantics, reference, iscaught up with intentionality and the use of language, and given the fact (to putit in his words) that “general issues of intentionality, including those of languageuse, cannot reasonably be assumed to fall under naturalistic inquiry” (2000a:132, 45), then it should go into a domain which unlike syntax is avowedlynon-naturalistic in the descriptions and explanations it gets: pragmatics.

Philosophers have tended to make the contrast between pragmatics andsemantics rest on the distinction not between those areas of language whereintentions are and are not involved respectively, but rather between those areaswhere non-linguistic intentions are involved and those where linguistic inten-tions are. For philosophers, notions such as reference and truth-conditions whichgovern semantics need not eschew intentions. After all, one may have an inten-tion to use a sentence with certain truth-conditions. This would be a linguisticintention, unlike an intention to use a sentence to get people to believe some-thing, or do something, etc. For philosophers, it is only the latter which falloutside of semantics and in pragmatics. But it is a mark of Chomsky’s deepcommitments to a scientific and naturalistic understanding of linguistics thathe allows it (and nothing else) to drive his basic classificatory criteria of thevarious areas of the study of language. Since for Chomsky intentions of anykind are unsuitable for a scientific and naturalistic treatment, the philosopher’sattempt to distinguish semantics from pragmatics by appeal to two notions ofintention misses the mark. They both fall in pragmatics, and all the rest is syntax,which is now (compensatingly as a result of the narrowing of linguistics to onlytwo broad areas) itself to be thought of more broadly than philosophers havethought of it, to include some areas of a naturalistically and internalisticallytreatable semantics in which no notions of reference or of intentionality occurat all.6

Is there a mind–body problem?

That leaves us with the other remaining source of modesty listed at the openingof the chapter. It has to do with the fact that (1) we may fail to formulate clearquestions before trying to answer them, and (2) our cognitive capacities areessentially limited and, so, some of the questions we formulate may not be oneswe could ever answer. The second kind of question, Chomsky calls “mysteries”and cites the problem of human freedom as an example; and he cites themind–body problem as an example of the former. Chomsky himself does not



say very much about mysteries, except to see them as a direct result of ourcreaturely limitations. Some philosophers have run together mysteries with ill-formulated questions, and the discussion in this section will try to give a detailedexposition of Chomsky’s views and disentangle the issues involved in (1)and (2).

As stated, a central example of this form of modesty can be found inChomsky’s discussion of the so-called “mind–body problem” which, he claims,can no longer be so much as formulated. This claim is bound to meet with resis-tance from some philosophers who think they have formulated a mind–bodyproblem and, indeed, regard their formulation as a central issue in philosophy ofmind. But when we view their problem from the perspective of Chomsky’s skep-ticism about its formulatability, it emerges that the problem is not best regardedas a mind–body problem at all. Their problem is really generated by what theyclaim is the sui generis nature of consciousness and this is as much a mind–mind problem as it is a mind–body problem. This leaves Chomsky’s centralpoint about the unformulatability of the mind–body problem standing. Further-more, it’s not clear what to say about the problem concerning consciousnessthat these philosophers think of as their mind–body problem. Colin McGinn(1994a), for instance, has argued that this problem is a mystery in Chomsky’s(1975, 2000a) technical sense – and many philosophers are likely to find thisplausible. If McGinn is right, then this is not an example of an obstacle to inquiryissuing from an ill-formulated question. Rather, we have a well-formulatedquestion that we cannot answer due to our cognitive limitations – what Chom-sky calls a “mystery.” The rest of this section shows why, from Chomsky’s pointof view, that is not a sensible conclusion. For, from his point of view, it wouldrequire that we have a clearer and more positive conception of consciousnessthan we do. Hence, like the mind–body problem, the problem of consciousnessis also best set aside until it too admits of clearer formulation.

Why does Chomsky claim that we cannot so much as formulate a mind–body problem? His reasons are both historical and scientific. According to him,there was a problem that Descartes could and did formulate. But our scientificcommitments have changed since then, and it is no longer possible to formulateone.7

Potted history of philosophy generally locates the source of Descartes’smind–body problem in his dualistic conception of the mind as an immate-rial and extensionless substance. But, according to Chomsky, the real source ofhis problem was his conception of body as a material substance subject to themechanistic thesis that causal influence on material bodies requires direct bodilycontact in space (contact mechanics). This is what precluded mental influenceover bodily events – not the immateriality of the mind but the idea that theonly way a mind could exert causal influence over bodily events would bethrough direct bodily contact in space. Chomsky observes that this mechanistic



conception of causation was contradicted by Newton’s theory of gravitationalattraction. Newton conceived gravitational attraction as a species of action at adistance which does not involve direct bodily contact in space – though, appar-ently, he did so with reluctance and even intellectual embarrassment. Oncemotion was reconceived along Newtonian lines, the mechanistic conception ofmaterial substance gave way to a much richer and open-ended conception ofmatter as capable, in principle, of supporting all sorts of properties that couldn’tbe counted as “material” in the sense that went together with Descartes’s contactmechanics. The ultimate consequence of Newton’s contribution, then, was tothink of the world as having many “aspects,” among which are mental aspects,and they are all properties of organized whatever-there-is. We can persist in call-ing this matter, but it is no longer something that stands in interesting contrastto the mental. So, Descartes’s problem vanished with the onset of Newtonianphysics. However, no one seems to have noticed. And, in consequence, noone has appreciated a more general lesson that Chomsky wants to draw fromDescartes and Newton, which is that it is no longer possible to formulate amind–body problem at all.

Let’s apply Chomsky’s lesson to what some contemporary philosophers taketo be their mind–body problem. They locate the source of their problem in themind – specifically, in the qualitative aspects of consciousness which Nageltried to capture with the phrase “what it’s like.” The problem is supposed to bethat it is impossible to put consciousness in this sense into perspicuous relationto physical nature. Thus, these philosophers tend to go along with potted historyand locate the real source of Descartes’s problem about mental–physical causa-tion on the mental side as well, in his conception of the mind as an unextendedsubstance. They focus so exclusively on Descartes’s metaphysical conceptionof the mind as an extensionless substance that they overlook the crucial rolethat his mechanistic conception of body played in generating his problem. Thishas led them to overlook, in turn, how changes in the notion of body that werewrought by Newtonian physics dissolved Descartes’s problem. Moreover, byignoring these historical facts, we ignore the more general fact that our physi-cal concepts must always play a central role in generating any serious problemabout the relation between mind and body. This is highly significant, becauseour physical concepts have not proved to be very stable. If they are not stable,then none of the alleged problems raised about the mind–body relation can bestable either. Inevitably, any new version of the problem will, like Descartes’soriginal, employ specific physical assumptions. And, so, like Descartes’s, ittoo may dissolve if there are major developments in the physical sciences.According to Chomsky, this makes it impossible for philosophers to formulatetheir alleged problem about the relation of mind and body. For they aim toformulate a perennial problem that is not going to be either solved or dissolvedby the future course of science. But, in order to formulate such a perennial



problem, philosophers would have to have two well-defined terms – mind andbody – such that we could see a permanent difficulty about understanding theirrelation. In Chomsky’s view, the latter term simply isn’t well defined. It hasproved to be a plastic notion that has undergone a succession of profound alter-ations and developments in the course of the history of the physical sciences.Furthermore, Chomsky holds that it is both historically inaccurate and method-ologically unreasonable to concentrate so much on alleged contrasts betweenthe mental and the physical per se. The best way to view the problem is in termsof the extent to which various aspects of the world are unified – where suchaspects include not only the physical and the mental but, also, the optical, theelectromagnetic, the chemical, etc.

Some might wonder whether current physical theory provides a way to for-mulate at least a current, if not a perennial, problem analogous to the problemthat mechanistic physical theory generated for Descartes. Here is a crude ges-ture in that direction. On one interpretation of contemporary physical theory,physics is complete in the following sense: it tells us that everything in theworld is constituted of particles and energy, and it specifies all of the laws bywhich all possible changes in particles and energy ever take place. This leavesno causal role for specifically mental phenomena to play in the physical world.And the same goes for any other phenomena that might be posited by anyspecial science. The point is not that there shouldn’t be any special sciences,or that there is nothing for them to do. They can, as they always have done,try to identify and explain various phenomena. The point is that if they suc-ceed in identifying and explaining real phenomena, both the explananda andexplanantia must ultimately be physical. This is directly implied by the com-pleteness of physics. Therefore, if there are any aspects of mind that we believecould never be understood in this way – as physical phenomena wholly gov-erned by physical forces – we have a current version of the mind–body problemafter all. It is a problem about how to reconcile our conviction that currentphysical theory is true with our conviction that irreducible mental phenomenaare real.

This interpretation of contemporary physics, which holds that all of the spe-cial sciences should ultimately be reducible to physics, is controversial amongphysicists themselves, and Chomsky doesn’t endorse it. His view is that the spe-cial sciences should be free to proceed autonomously, leaving it open whetherwe will or won’t be able to account for their results within physical theory.Although we are sometimes able to provide such an account, there is no guar-antee that we will be able to in every case. And Chomsky insists that evenwhen we can, the result may fall well short of reduction. He prefers to describesuccess in this domain as the successful unification of a special science withphysics. There is no guarantee that such unification can be achieved in everycase, any more than reduction. And even when such unification can eventually



be achieved, it may have to wait until after our physical concepts themselveshave undergone substantial change – as they had to before chemistry could beunified with physics. Therefore, it is always wrong in Chomsky’s view to inhibitthe course of the special sciences with the requirement that we be able to finda way to account for their findings within current physical theory. Yet it wasby imposing exactly that requirement that we managed above to formulate acurrent version of the mind–body problem analogous to Descartes’s original.Chomsky would not be moved. The general lesson he draws from Descartes andNewton stands: the instability of physical concepts should make us sensitive tothe difficulty of formulating a serious mind–body problem.

Let’s consider more carefully the standard formulation that philosopherstry to give. We’ve already indicated that, for them, the problem arises fromthe sui generis nature of consciousness – in particular, its qualitative aspects.The problem is that any particular physical facts can always be imagined tobe the same while the qualitative aspects of consciousness are imagined to bedifferent or even, perhaps, altogether absent. (Note that in this formulation,the term “physical” does not refer merely to the subject-matter of physics,but is intended to refer more broadly to a whole range of phenomena whichincludes chemistry and neurophysiology.) Chomsky’s quick reply would be thatwe won’t have a well-formulated mind–body problem until we’ve been given acoherent account of both terms of the problem. But the philosophers who claimto find a problem here don’t agree, because they believe that it will persist inthe face of any evolution of our physical (and here, again, they use this term inthe broadest sense) concepts. This is a point to which we will return. But first,a digression about philosophers who, like Chomsky, deny that there is any suchphilosophical problem. We register some important differences between themand Chomsky.

Certain philosophers – call them the optimistic materialists – are unimpressedby the feats of imagination that allegedly generate the philosophical problem(imagining the qualitative facts of consciousness altered while the physical factsremain the same). They see this as nothing more than a reflection of our tempo-rary ignorance about the physical basis of consciousness. And they confidentlypredict that we will eventually overcome our ignorance. In their view, once wehave discovered the physical basis of consciousness, we will find it perfectlynatural to identify conscious phenomena with physical phenomena – just as wehave come to find it natural to identify heat with molecular motion. And thiscan be so even if we retain the capacity to imagine them coming apart (afterall, we continue to be able to imagine defying gravity even though we havecome to believe a theory according to which it is impossible). In addition tothese optimistic materialists, some dualists might also agree with Chomsky thatthere is no mind–body problem. The fact that we can’t put consciousness intoperspicuous relation to the physical world is merely a sign of their distinctness.



It is important to see that, when Chomsky denies that there is a mind–bodyproblem, he isn’t aligning himself with either of these philosophical positions.He is not taking up a metaphysical position like reductive physicalism or dual-ism. He regards such metaphysical issues as scientific issues that haven’t beensufficiently sorted out to be a basis for any significant conclusions about mindor body or their relations. His opposition to the idea that there is a problemsprings, as noted, from modesty about whether we can actually formulate aclear problem.

What, then, of the philosophers who think they can do this? Colin McGinnis a particularly fitting example since his account of the mind–body problemis inspired by Chomsky. In fact, he sees all philosophical problems as myster-ies in Chomsky’s technical sense. It is philosophy’s fate as a discipline to becondemned to mysteries as its subject-matter.

A little more needs to be said to expound Chomsky’s (1975) notion of a mys-tery. A mystery is a problem that we can’t solve due to our intrinsic cognitivelimitations. In Chomsky’s view, we are bound to be subject to such limitations.This is because our cognitive capacities are, au fond, biological, and any bio-logical capacity is bound to have limitations. Furthermore, such capacities andtheir limitations are bound to vary from species to species. Take, for example,the respiratory capacities of different animals. Mammals absorb oxygen intotheir systems from the air they breathe. The respiratory capacity by which mam-mals do this enables them to live wherever there is air to breathe and preventsthem from living anywhere there isn’t. They cannot live underwater as fish do.Likewise, the particular respiratory capacity by which fish absorb oxygen fromwater simultaneously enables them to live underwater and prevents them fromliving out of water. They cannot live in the open air as mammals do. Chomskyapplies the same point to cognitive capacities. If an animal is cognitively con-stituted in such a way as to be able to solve certain sorts of problems, thatautomatically makes the animal ill-equipped to solve other sorts. The prob-lem that has interested Chomsky the most is the one that children confront inacquiring a human language (“Plato’s Problem”). Children are able to solve thisproblem because each human language is an instantiation of a single underlyinguniversal grammar, and we have innate knowledge of that grammar as well as acapacity to identify instantiations of it. Chomsky speculates that once in place,the language capacity enables us to solve a whole range of problems that aresoluble via the same sorts of iterative and recursive structures that characterizehuman languages – a central and important case being mathematical problems.Other species, such as wasps, obviously lack human language and, hence, can-not deal with the whole range of problems that the capacity enables us to solve.We should not infer that wasps are mindless. The cognitive capacities of waspsare simply different from ours, and enable them to solve the specifically waspishproblems that they need to solve. Similarly, there could be other species who



stand to us as we stand to wasps – species whose members have cognitivecapacities by which they can solve a whole range of problems that we don’thave the capacity to solve.

Typically, the cognitive problems that the members of a given species can’tsolve will lie entirely beyond their comprehension. So, not only is it the case thatwasps cannot solve mathematical problems; they cannot cognize the problemsto begin with. But Chomsky doesn’t think it has to be this way, especially notin the human case. He finds it plausible that we human beings actually havethe ability to formulate some of the problems that it is beyond our cognitivepowers to solve. These problems which we can formulate, but whose solutionslie beyond our cognitive grasp, are mysteries for us. (Though they might besoluble by species with different cognitive capacities.)

McGinn agrees with Chomsky that cognitive capacities are bound to bringin train cognitive limitations, with the consequence that some of the problemswe face may not be soluble by us – and, hence, mysteries in Chomsky’s sense.According to McGinn, these mysteries are the problems that characteristicallyengage philosophers. His point is not that philosophers have never confrontedproblems human beings are able to solve. Rather, when a problem becomestractable, it ceases to be a specifically philosophical problem and becomesinstead a topic of scientific investigation. Thus, the real and proper domain ofphilosophy is the class of perennial problems that human beings can never solveand will always be perplexed by. This way of thinking about philosophical prob-lems is a naturalized descendant of Kant’s critical philosophy, a fact to whichMcGinn expressly draws attention with the label “Transcendental Naturalism.”McGinn diverges from Kant in that he sees no reason to suppose that we havea priori insight into our cognitive constitution and its built-in limitations. Inthe place of Kant’s a priori investigation, he offers an empirical speculation.He speculates that a problem can be solved by human beings only if it can besolved by invoking two related cognitive strategies, namely, Combining Atomsand devising Lawlike Mappings – CALM for short. This speculation is partlymotivated by some suggestions of Chomsky’s about how our language capac-ity might have acquired wider cognitive application. But it can’t be said thatwe have very much direct psychological evidence that our cognitive abilitiesare completely circumscribed by the limits of CALM thinking. And McGinnrecognizes this. He draws other, less direct evidence for his speculation fromthe history of philosophy. He asks us: which of the problems that have engagedphilosophers eventually became scientifically tractable? Which have continuedto resist scientific understanding and, as a result, continued to perplex us in theway that is characteristic of a truly philosophical problem? In his estimation,all the problems that proved to be scientifically tractable yielded to CALMstrategies, whereas all the problems that have continued to perplex philoso-phers didn’t. He infers that CALM strategies really do circumscribe the limits



of human understanding. And, like Kant, he recommends that philosophersshould recognize these limits and stop trying to answer the questions that can-not be answered from within them. If McGinn is right, we face a new end ofphilosophy.

It stands to reason that Chomsky would be sympathetic to McGinn’s accountof philosophy.8 The general spirit of the account is Chomskyan, embracing hisnaturalistic conception of our cognitive capacities, along with the corollary thatthere are bound to be limits on human understanding. And Chomsky seemsto find truth in McGinn’s picture of the characteristic activity of philosophers,which is to wallow in insoluble mysteries rather than get on with the constructivebusiness of science.

However, his sympathy with McGinn must stay at the level of generalities.For one thing, McGinn’s speculation about CALM is insufficiently supportedto win unqualified endorsement from a working scientist like himself. But themore important dividing issue concerns one of the uses to which McGinn putsthe speculation. He uses it in order to formulate the very problem about mindand body that Chomsky claims can no longer be formulated. CALM, remember,is supposed to distinguish the soluble problems from the insoluble, where suchinsolubility is the mark of the truly philosophical problem. And McGinn regardsthe mind–body problem as a paradigmatic example of the latter. In fact, he holdsthat CALM provides a particularly good way to formulate the problem. Theproblem is that we can’t account for consciousness in CALM terms, as arisingin the material world due to the ways in which physical atoms are combined ordue to lawlike mappings among physical phenomena.

It is odd that McGinn should focus on the mind–body problem, given thathis overall account of philosophical problems derives so much of its inspirationfrom Chomsky, and Chomsky goes to such pains to argue there is no suchproblem. It is even odder that McGinn never addresses this argument. Chomskycites McGinn’s overall account of philosophical problems as mysteries in hissense, but he too passes over in silence this difference between them. Althoughboth of them would clearly prefer to emphasize what they have in common, itproves instructive to sort out this difference between them.

How might McGinn respond to Chomsky on the mind–body problem? SinceChomsky’s official position is that we can no longer so much as formulate it,McGinn might simply point out that he has formulated it, in terms of CALM. Hemight buttress this response with the following line of thought: when Chomskydenies that there is a mind–body problem, his main concern is to liberate themental (and other special) sciences from any theoretical constraints that mightderive from the current state of physical theory, his thought being that oth-erwise we might fall into mistakes similar to the one Descartes made whenhe thought he needed to account for mind–body interaction within the con-straints of mechanistic theory; however, although it is sensible and laudable



that Chomsky should thus try to spare future mental science from similar mis-takes, his argument doesn’t touch the version of the mind–body problem thatderives from CALM: CALM is not proposed as a transient constraint that hap-pens to derive from some current theoretical commitment. CALM is proposedas an insuperable constraint on all human thought.

Its hard to say anything in detail about how Chomsky would evaluate thislast, highly speculative, proposal of McGinn’s, except to repeat what was saidabove about how little direct empirical support it has. But it should be regis-tered that the speculation seems, on the face of it, wholly non-credible. It seemsobvious that it is within our power to conceive physical phenomena in non-CALM terms, specifically, in teleological terms. We do not have in mind the“as if” teleology often invoked in ordinary descriptions of biological phenom-ena, as when one says that a tree’s roots grow downward in order to find wateror that the heart pumps blood in order to send nutrients to different parts of thebody. Everyone knows that roots and hearts don’t really have ends. Nevertheless,what we now take ourselves to know about this doesn’t exhaust what we canconceive. We can (whether plausibly or not) without incoherence conceive thatall sorts of physical things might really possess ends. Aristotle certainly did.Moreover, we can conceive this in a non-CALM way. Let’s offer an explanationby way of contrast with a CALM conception of what we are calling “as if” ends.Conceived in CALM terms, the possession of an end is the possession of a dis-position or lawlike tendency to achieve the end. Call ends in this sense “as if”ends because they are reducible to something that can be described in perfectlynon-teleological terms, namely, laws. But now consider minded things. Whenwe conceive minded things as having ends, we don’t necessarily conceive themin CALM terms, as having dispositions or lawlike tendencies to achieve theirends. We can conceive them instead as merely striving to achieve their endsin a very limited sense that falls short of a lawlike tendency. This is what wemean by a genuine teleological conception, as opposed to an “as if” one – aconception in which behavior is directed towards ends in a way that cannot bereduced to a disposition or lawlike tendency. Of course, the only domain inwhich we currently find such a non-CALM teleological conception natural andplausible is the intentional activities of minded things. (See the first section ofthis chapter for a more extended discussion of why intentional activity is neitherdispositional nor lawlike.) But the point is that this is not because it is some-how beyond our cognitive powers to apply the conception to material thingsas well.

It is worth registering this point for two reasons. First, it brings out thatMcGinn’s speculation about CALM might turn out to be false. CALM might notbe an insuperable cognitive limitation, but a theoretical constraint that derivesfrom our current theoretical commitment to giving non-teleological accountsof natural phenomena. And this bears directly on the question whether McGinn



has successfully escaped the critical force of Chomsky’s argument against thosewho attempt to formulate a contemporary mind–body problem. If CALM isa mere theoretical constraint that derives from the current state of scientificunderstanding, then Chomsky is within his rights to suggest that we ignore it –just as Descartes would have done well to ignore the problem that mechanismposed for his account of the mind, on account of the fact that Newton eventuallyprovided an alternative physical theory in which the problem didn’t arise. Thisis not to say that Descartes was in a position to know that this would eventuate.Chomsky’s point is that we have the benefit of hindsight and shouldn’t fall intothe same trap.

At this point, McGinn might retreat to a less ambitious position from whichhe could still try to formulate a serious mind–body problem. For, even if itcan’t be shown that CALM is an insuperable cognitive limitation, CALMapproaches have proved to be a real scientific advance over earlier teleologicalapproaches. So, McGinn might reason, it is a good bet that CALM is here tostay, and yet consciousness could never be accounted for in CALM terms. Thatcounts as a serious problem, Chomsky’s lessons from Descartes and Newtonnotwithstanding.

This brings us to the second reason for raising the possibility of a non-CALMteleological conception. When McGinn was confronted with this possibility, hedidn’t say he found it inconceivable (even though this is entailed by his ownspeculation about the limits of our understanding). He was prepared to allow,at least for the sake of argument, that it is conceivable. And he didn’t say whatmost of us these days believe, which is that non-CALM conceptions are notexplanatory and, so, will never have a place in a scientific account of the relationof mind and body. Instead, he insisted that the mind–body problem would ariseeven within a non-CALM teleological framework. According to him, we canno more understand how consciousness could arise out of non-CALM teleolog-ical processes than we can understand how it arises from CALM phenomena(McGinn 1994b). This means that McGinn misled us when he stated that CALMprovides a good way to formulate the mind–body problem. But the aim hereis not to lodge that complaint. The aim is to raise the following question: whatalternative formulation can McGinn give, which makes clear what there is incommon between CALM conceptions and non-CALM teleological concep-tions, and which also makes clear why it is impossible to put the concept ofconsciousness into intelligible relation to either one? It turns out that he, likemost contemporary philosophers, does not appeal to any general features ofbody (or the material and physical more generally), but appeals rather to spe-cial features of consciousness – in particular, to the idea that the qualitativeaspect of consciousness (“what it is like”) is sui generis. Of all the things in theworld, this is the only thing that can be known only from a first-person point ofview. Everything else can be known and conceived from a third-person point of



view. And it is this dichotomy between the first person and the third that reallymakes for the mind–body problem as it is typically understood by contemporaryphilosophers, including McGinn. The problem arises because we can alwaysimagine the phenomenological facts that we know from a first-person point ofview being different or absent, while all of the other facts that we know from athird-person point of view remain the same.

It is understandable that many philosophers would be inclined to think thatthis formulation of the mind–body problem is untouched by Chomsky’s argu-ment to the effect that there is no such problem. He may be right that ourphysical concepts may change in unanticipated ways. He may also be right thatthe special sciences, including the mental sciences, should not be constrainedby the commitments of current physical theory. But, all the same, our physicalconcepts will never cease to be concepts of things known from a third-personpoint of view and, in consequence, the concept of consciousness can never beput into perspicuous relation to them. Those who take this view of the mind–body problem, as arising from the first-person/third person dichotomy, may besomewhat incredulous that Chomsky doesn’t see the problem.

Chomsky might point out that the conception of something as knowablefrom the third-person point of view is not a specifically physical conceptionat all. Many mental facts are also knowable from the third-person point ofview – for example, facts about intentional attitudes, action and perception.And qualia pose the very same problem in connection with these mental factsthat they pose in connection with physical facts. That is, we can always imaginethat all of the publicly knowable mental facts remain constant while the factsconcerning the qualitative aspects of mental life are permuted or absent – as inthe much discussed cases of inverted spectra and zombies. These possibilitiesfollow from the essentially private nature of qualia, as things knowable onlyfrom a first-person point of view. And that is not all. If qualia truly are private,then they are epiphenomenal. For, if they had any systematic effects in therest of mental life, that would make them knowable from a third-person point ofview – they would not be private. Wittgenstein concluded: so much the worse forprivacy. But contemporary philosophers who recognize a mind–body problemdo not concur. They are prepared to acknowledge that qualia are known onlyfrom a first-person point of view. That is why they think they face a mind–body problem. But all this brings out, on behalf of Chomsky, that this problemis not well described as a mind–body problem. It’s just as much a mind–mindproblem – a problem about how to understand the relation of qualia to the rest ofmental life. So, if McGinn is right to see consciousness as generating a mystery,it shouldn’t be regarded as a mystery about the relation of consciousness to bodyper se.

Most philosophers who recognize the qualitative aspect of consciousnesswould recognize that it generates a mind–mind problem. But they wouldn’t



see much point in denying that there is a mind–body problem. In fact, theywould very likely view both problems as special cases of a single, more generalproblem concerning the relation of the qualitative aspects of consciousnessto everything which is public, where that includes all aspects of the physicalworld as well as the public aspects of mind. Chomsky, by contrast, wouldn’tsee much point in lumping everything public together in this way. That makes itseem as if the pressure to find some interesting relation between the qualitativeaspects of consciousness and physical facts is as great as the pressure to findsome interesting relation between them and other mental facts. But that justisn’t so, according to Chomsky. It might seem to be so if we had a physical-ist bias according to which everything must ultimately be related to physicalfacts. But that is a bias. We ought not, in Chomsky’s view, to prejudge howmuch unity we will eventually achieve within the sciences. For all we know,the mental sciences might always remain quite independent of the physicalsciences. If things were to turn out that way, then no mental phenomena couldbe put into interesting and perspicuous relation to the physical facts. And, inthat case, it would obviously be completely inappropriate to ask that the qual-itative aspects of consciousness, in particular, be put into such relation. Butthis would not constitute a problem. It would just be how the world turned outto be.

It might appear that the mind–mind problem still stands and remains pressing.For, from the first-person point of view of consciousness, the fact that there issomething it is like seems to go to the heart of our conviction that we are minded.This conviction runs so deep that it leads many neo-Cartesians to equate beingminded with being the subject of conscious states that have a qualitative aspect.As soon as the equation is made, the mind–mind problem seems to disappear.The role of the qualitative aspects of consciousness in the rest of mental lifecomes to seem utterly straightforward: they are constitutive of the whole domainof the mental. To view the mind in this way is to see it as essentially private.What we have been describing as the public aspects of mind are not aspectsof mind at all. They are mere evidence of mentality, the actual existence ofwhich can be settled only from a first-person point of view. The neo-Cartesianconception therefore licenses skepticism about other minds: the limitation onour knowledge of what it’s like for others is a limitation on our knowledge ofother minds altogether.

Chomsky opposes the neo-Cartesian conception of the mind on numerousgrounds. He opposes any attempt to circumscribe the mind by any single crite-rion, such as consciousness or intentionality. And he finds the criterion of con-sciousness particularly implausible. It would entail that linguistic knowledgeisn’t really knowledge (on the ground that we are not conscious of it) – some-thing he emphatically denies. Furthermore, it would make any science of themind impossible. There are also familiar philosophical grounds of opposition



issuing from Wittgensteinian considerations. The tie between intentional factsand publicly knowable behavioral facts is closer and deeper than neo-Cartesiansallow – indeed so close as to rule out privacy altogether. From this Wittgen-steinian perspective, there is neither a mind–mind problem nor a mind–bodyproblem – at least not of the sort we have been discussing.

Perhaps it would be going too far to dismiss privacy altogether. We can allowthat there is something it is like to feel a toothache or see peacock green which isknowable only from a first-person point of view, where that brings in train somereasonable doubts about whether we can really know what it is like for others,and even brings in train the less reasonable but nevertheless coherent worry thatin some cases there may be nothing it is like. We can do this while insisting thatanalogous doubts and worries about our knowledge of related intentional facts,such as the desire to visit the dentist or the belief that there are peacocks in thevicinity, are misplaced. If we allow and insist upon all this, we are concedingwhat McGinn and other philosophers take to be the source of their problem.But it must be stressed that if any problem could be formulated here, it wouldbe a mind–mind rather than a mind–body problem. And it is highly doubtfulthat Chomsky would regard it as a clearly formulated problem, given the forceof the Wittgensteinian considerations against the very idea of privacy.

It has been explained in some detail why Chomsky thinks it is (now) impossi-ble to formulate a mind–body problem. The reasons are historical and scientific.His overall views would also give him reason to doubt whether we can formu-late a coherent mind–mind problem. Here the reasons are less historical andscientific, and closer to philosophical reasons of the familiar Wittgensteiniansort. Insofar as the alleged problem is supposed to be generated by the factthat there are qualitative aspects of consciousness which are essentially private,there is nothing we can do but apprehend them from our own first-person pointsof view and, otherwise, remain silent about them. In doing this, we wouldn’thave formulated a problem at all. That is the third reason to be modest aboutthe scope of inquiry.


10 Meaning and creativity

James McGilvray

Introduction

Often, philosophers, linguists, and cognitive scientists aiming to construct atheory of meaning for languages focus on how words are used by people to dealwith the world. They might assume there is a regular way in which the worddolphin is used – that people regularly use dolphin to refer to or denote a class ofaquatic beasts. And they might assume that there is a central, core use of words –using She has a pet dolphin to correctly describe some state of affairs, perhaps.If these assumptions were correct, dolphin might then be defined in terms ofsome regular function(s) it serves in a community of individuals who “speakthe same language,” understood as “use words in the same way” (cp. Davidson1967, Sellars 1974; contrast Chomsky 1996a, b, 2000a, Fodor 1998). Theseassumptions are built into technical terms: “truth (or correctness) conditions,”“functional (or conceptual) role,” etc. Because these attempts focus on commu-nities, circumstances, things, and so on, I call them “externalist” approaches.

Externalist assumptions are wrong. Chomsky (1959, 1966, 1975, 1980,1981a, 1986, 1995b, 1996a,b, 2000a), recalling observations that go back toDescartes (see below), repeatedly points this out. Words do sometimes get usedby people in similar ways; and they do sometimes come to be related to theworld. But they do neither by themselves. Similar uses and relations to the worldare products of human actions, of words’ free and typically creative use byhumans. Because people use words for all sorts of purposes, because the use oflanguage is a form of free action, and because there is little reason to think thatthere can be a science of free action, there is little reason to think that there canbe a naturalistic externalist theory of meaning.

The key to a naturalistic science of meaning for human languages lies inadopting an internalist strategy: look inside the head. As other contributorsto this volume show, the language organ, an innate biological faculty, pro-duces expressions (sentences) consisting of sounds and meanings. Think ofthe meanings of sentences as language’s conceptual contributions to humancognitive capacities and activities – including referring, inferring, and sayingsomething true. Think, then, of linguistic meanings as a potentially infinite set of

204


Meaning and creativity 205

internally constituted sentential concepts. These are theoretically defined men-tal entities that “interface” with other mental systems and that – as we put itin ordinary speech – are used by people to do various jobs. Chomsky infor-mally calls them “meanings” or “perspectives”; formally, they are “LF”s and –recently – “SEM”s (for “semantic interfaces”). Naturalistic sciences of them –thus, of meanings – are in place.

Abandoning externalist assumptions, we also gain insights into human free-dom and creativity. Language, offering us internally constituted conceptualmaterial unavailable to other creatures, turns out to be an extremely usefulorgan. It offers readily accessible conceptual resources to deal with the myriadcognitive tasks of getting along with others and understanding the world. Itscontributions begin early; it is automatically acquired at an early age and, withits resources available, the child easily develops common sense (folk physics,folk psychology, etc.) and engages in play and fantasy. Its continuing availabil-ity and increasingly sophisticated use in solving all sorts of everyday problemscontribute to the cognitive functioning of adults, enabling creative solutions tocognitive tasks as varied as those that confront citizens, gossips, engineers, andartists. So by looking inside the head and not at the extraordinary diversity ofways language is used, the scientist of meaning can stop trying to constructa theory of language use, and the philosopher of mind and interested laymancan begin to appreciate how a biological organ plays a crucial role in offeringhumans the individual and cultural diversity and creativity that are character-istic of the human species. Abandoning the quixotic task of constructing ascience of linguistic action and behavior, an internalist and rationalist approachto language and mind makes sense of how language can be so useful, and whyhumans have such distinctive mental powers.1

A word on why this chapter is where it is. Because meanings are defined byappeal to Universal Grammar (UG), a discussion of linguistic meaning alonecould be placed in the linguistics section. But because linguistic meaningscontribute so much to human creativity and because the way in which they dosuits only an internalist and rationalist picture of the mind, it also belongs in asection on Chomsky’s view of the human mind. Finally, because language withits unlimited number of concepts is a biological engine that provides for andeven encourages a distinctively human form of creativity, and because creativityand the need for it play such important roles in Chomsky’s political views, achapter on meaning and creativity rightly appears immediately before a sectiondealing with his political views.

Defining linguistic concepts

If linguistic meanings at the interface with other systems are biologically con-stituted sentential “perspectives,” how does one define them? The meanings of


206 James McGilvray

sentences are composed from the meanings of their words. So we can begin byasking what the meaning of a word is.

Ordinary dictionaries are little help. They do not define the meanings ofwords – the relevant “atomic” and/or complex concepts (Chomsky 1996a, b: 24).They assume that a person already has a natural language and access to linguis-tically expressible concepts. For those who do, dictionaries might do two things.First, Webster’s and the Oxford might inform a person about what sounds peoplein a population associate with specific meanings. While possibly useful, thisis irrelevant to a naturalistic science of meaning. Meaning–sound associationsare arbitrary; there is no biological relationship between them. The conceptwash could be as easily associated with the sound “wiggle.” Associationsreflect what Chomsky calls “Saussurean arbitrariness.” These associations aresocial conventions, of no interest to the natural scientist.

Second, dictionaries might provide hints and prompts. They hint by listingin their “definitions” a few terms that point in the direction of a concept. Theyprompt by proffering a context in which a word is used. They display a phrasecontaining words that, except for the target one, are likely to be familiar; theseallow the speaker to create a context of use for the unfamiliar word and producein their heads a concept that they might use in that context. The full-fledgedOxford offers hundreds of thousands of examples of both techniques. Dependingon whether a person has already positioned a concept in their mental dictionaryor not (i.e. associated a concept with a specific sound and perhaps come to usethe associated elements), a hint or prompt might remind or trigger. Remindingneeds no explanation. Triggering is a matter of some prompt generating orproducing the concept in one’s head; I return to it in a moment. So far, theimportant thing to notice is that dictionaries work only because people eitheralready have in their active mental dictionaries, or can readily insert into one, therelevant concepts. They do not define; they hint or prompt, leading to remindingor triggering.

Assume for the moment that words as the naturalistically inclined linguistunderstands them consist of sounds and concepts. A scientific definition of aword must consist, then, in a full theoretical specification of a sound and ameaning. Defining sounds is a job for phonologists and phoneticians (Dresher,this volume). While defining sounds is a fascinating task and there are suggestiveparallels with defining meanings/concepts, this is not the occasion to attempteither. The scientist of meaning aims at individual (root) concepts and howthese are put together to make sentential concepts. Linguists already know alot about how the mind puts sentences together, for there has been considerableprogress in syntax and morphology – the sciences that deal with how. Therehas been less progress in getting a grip on individual concepts. But the strategyis clear: as with phonology, syntax, phonetics, and morphology, look inside thehead.



One reason to look inside, mentioned above, is that language use is out ofreach of science, for use is creative. Another is found in poverty-of-stimulusfacts. They apply to sounds and meanings too. Over the lifetime of an individual,a person easily acquires thousands of linguistically expressed concepts, eachacquisition an example of poverty of the stimulus. Children, given little data,acquire approximately a (root) word a waking hour between the ages of two andeight (Chomsky 1995b). The capacity to activate concepts and quickly asso-ciate them with sounds (“learn words”) continues into later life. Unlike the fullrange of linguistic sounds,2 there do not seem to be temporal barriers (closed“windows of opportunity”) to swift acquisition of the concepts expressed innatural languages. While it seems children need to acquire some before others(dog before chihuahua), there are no obvious parametric switches set soonafter birth that channel the language organ’s efforts in a particular direction,excluding others. For concepts, the seventeenth-century Cambridge PlatonistRalph Cudworth spoke of circumstances “occasioning” and “inviting” a con-cept. To make this possible, he thought, (lexical) concepts must somehow beprefigured (anticipated – he called it “prolepsis”) in an “innate cognoscitivepower.” The nineteenth-century German linguist and philosopher Wilhelm vonHumboldt said the mind must have an innate productive “mental instrument”that yields lexical concepts: when prompted by a “signal” from a person’sspeech, “matching but not identical concepts are engendered” in the mind ofthe hearer (Humboldt 1999: 152; Chomsky 2002a: 102–3). If so, a naturalisticdefinition of a lexical concept consists in providing a theory of the relevantmechanisms – the innate machinery – that yield it and others.

Triggering, mental machinery, and theoretical definition

Although the poverty facts indicate that some natural – neither use- nor social-dependent – process is at work, no one really knows how triggering takes place.Some occasion, external or internal, “occasions” the mind’s machinery, and aconcept is activated. As indicated, the hints and prompts of dictionaries can servethe purpose: similar concepts, or a context of use, trigger a concept and associateit with a sound in a person’s mental dictionary. Overheard conversations serve astriggers, as do novels, the efforts of others to inform, poetry readings, pictures,etc. What philosophers call “ostensive definition” – for example, pointing to ametal object with two handles and saying “spokeshave” – might do. Occasionswhere someone encounters something for the first time might too: a Filipinachild might acquire the concept snow when getting off a plane in Montreal inJanuary. She might not acquire exactly what another person has, and likely herconcept will undergo modifications later. Her uses of the concept will certainlybe variable; they likely will become more sophisticated too. But these are details.The important point is that she must have a mind biologically like that of others


208 James McGilvray

for snow to be triggered as readily as it is, and to be like others’ snow . Shemust have mental machinery devoted to language, plus machinery (it may bethe same) that assembles or packages specific concepts that can be insertedin linguistic derivations to produce sentential concepts. Whether specificallylinguistic or not, internal machinery sets the agenda and dictates what counts asa trigger: the mind with the relevant machinery “looks” for some pattern in theinput that sets it on a course it “anticipates” already. Laura Petitto’s discussion(this volume) of the onset of babbling provides a simple, important, example.Apparently, the language organ is not shaped by experience; it brings concept-forming machinery to experience. In this sense, individual concepts are innate,or latent in (the machinery of) our minds. If we are to define them, we must –as with sounds and signs – deal with that machinery.

To avoid confusion, keep in mind that to say that the concepts that typicallyappear in natural languages are anticipated in the machinery of the mind is not tosay that we are conscious of them, when latent or active. Nor is it to say that theyare “there” ready-formed, with all the features that constitute a specific lexicalitem already assembled in the form they take in a person’s working vocabulary.Certainly it is not to say that they come already linked to things or classes ofthings in the world. It is to say that they are products of biological machinerycapable on an occasion of constituting or “activating” them, yielding items withconfigurations that when placed at language’s interface with other systems canaffect human cognitive functioning, thereby affording a capacity to recognize,distinguish, gauge, assess, or otherwise use the concept. Perhaps a specificconcept is anticipated in something like the way the human immune systemanticipates a range of pathogens: the machinery provides for them. Clearly,like the immune system, the machinery that provides concepts to languagedoes not anticipate all concepts. Some concepts are native to other systems.Language cannot process most “qualia” (as philosophers call them), such ascolors. Language deals with the linguistic concept red , not with a (specific)visual red that people with a functioning visual system enjoy and that they“attribute” to the surface of an object outside the head.3 Scientific concepts,furthermore, unlike those that appear in our natural languages, are not virtuallybuilt into our biology. They are not easily acquired in the way the conceptsof natural languages are, but instead require sophisticated understanding of atheory and, typically, a lot of preparation and work.4 They seem to be created,or invented, by people who construct sciences. Chomsky holds that people havesome innate aid in constructing such theories: our “science forming capacity”(1975, 1988b) provides a kind of guidance. But the particle physicist’s conceptpion is not somehow anticipated in us at birth. If it were, the child wouldreadily acquire it.

Note that defining a concept such as paint in the way required for a sci-ence of language is not defining in the way philosophers traditionally assume.



Defining is not a matter of stating necessary and sufficient conditions on theapplication or use of a concept – stating under what conditions the transitiveverbal concept paint is true of an action performed by Vermeer or Sam thehouse painter. Mathematicians and scientists might sometimes in their workuse concepts such as aleph-null and pion in ways that are regular anddeterminate enough to give credence to the project of constructing such a def-inition (really, an idealized proposal for what counts as proper use among thepractitioners of the relevant science). But neither they nor anyone else usespaint in such a way. Thinking that this is what definition consists in missesa crucial point: the use of linguistic concepts is creative, and can be so justbecause natural language concepts and their internally provided “contents” areanticipated, or innate.

To define a linguistic concept naturalistically is to construct a scienceof Cudworth’s innate cognoscitive power or Humboldt’s mental instrument.Describing exactly how linguistic concepts are triggered and put together –describing the operations of our mental instruments – is a scientific project thathas only begun. Speculating, although not unreasonably, perhaps the mecha-nisms perform a kind of mental chemistry. Perhaps what is innate is a relativelysmall set of really primitive features and a much smaller set of configurationsinto which they can be put, plus some machinery that when prompted com-bines features in configured arrays – makes a structured compound, unique inits character. When a person is confronted by something that “needs” a conceptof a specific sort (for example, one that embodies what we understand whenwe understand refrigerator), the machinery provides it by inserting fea-tures in an array. This is a convenient picture of concept acquisition and hasthe advantage that it allows that the features that are put into configurations,and the configurations themselves, are smaller in number than the hundreds ofthousands (or more) root concepts that a human can acquire.5 The number ofacquirable concepts must be large. We must allow for those for which there isyet no need. The ancient Mesopotamian child did not need refrigerator ,even though s/he could easily have acquired it, given the relevant triggeringexperiences. Like most of us, however, s/he could not have easily acquired thetechnical concepts of the sciences that detail the principles by which refrigera-tion works.

Continuing to develop this picture, perhaps we can then think of morphologyor the operations of “making words” as adding structure and turning a rootconcept into an item of a specific lexical category (N, Vtr, A . . .). structbecomes the noun concept structure , the verb construct , and so on.Perhaps we can think of lexical categories as specific kinds of configurationsof features characteristic of the relevant category. If we include morphologicaloperations in the basic computational structure of the language faculty proper,and if as above we think of roots as assembled by internal machinery (which


210 James McGilvray

may or may not be a part of the core operations of the language faculty), we canthink of root concept and lexical category acquisition as being written into themachinery of the mind. It is a short step to thinking of the “features” that areassembled in root concepts and in morphological configurations (noun, verb,etc.) as “saying” what a concept and further structures added to it contribute tothe semantic interface – to sentential meanings. Finally – continuing the theme –we can think of a full theoretical specification of a sentential meaning as describ-ing the contribution that that sentence can make to any cognitive tasks to whichit might be put by further systems in the head and, eventually, by the person.

Examples give an idea of some of the features that must somehow cometo be included in a word’s concept. Verbs come in a set of forms that specifyargument places: a verb such as give can be thought of as an array that providesargument places for nouns assigned the features giver , receiver , given .Natural language noun features seem to be configured to answer Aristotle’s“causes” (explanations) of the things that the nouns can be used to refer to.They “say,” “this is its origin,” “this is what it does,” “this is what it’s made of,”and “this is its essence” (Chomsky 1975, 1995b, c; Moravcsik 1975; Pustejovsky1995).6 Take house : perhaps house’s features are something like artifact(“made by humans”), container (“envelopes”), rigid (“secure materials”),dwelling “holds humans.” In effect, a specification for the noun-concepthouse says what a speaker “knows” in Chomsky’s “know a language” sense –in effect, what a speaker with the concept house brings to his or her thoughtor discourse about houses.

Because this picture of natural language concepts, their acquisition, and theirstructural and “content” contributions to cognition is convenient and seemsconsistent with what is understood about the nature and structure of words andhow to develop a theory of them, I adopt it. It is at best a reasonable guess.7

And it is burdened with primitive terminology: no serious linguist expectsthat the terms for concepts and their defining features that I have been using(slug , artifact ) will continue to serve as theoretical vocabulary. We muststart somewhere, but – if we are on the right track – theory advances, and wemust expect changes and improvements, including a more articulate theoreticalterminology, which is likely to be as far removed from the starting point as anyadvanced science’s vocabulary and associated concepts are.

One reason such a vocabulary has not been easy to produce is that it mustbe integrated within a full and adequate theory of the language faculty. To seewhy, let us start with what might still be thought to be the simplest possibletask, defining a single “word” or lexical item. Let us assume that a theory ofmeaning for words says what lexical items are in terms of features. Assumetoo that the features in terms of which a definition will be given are those thatwill or can play a role in the derivation of a sentence. This is reasonable, forlexical items (or rather, their features) combine – in Chomsky’s terminology,



“Merge” – to form “sentences” or “expressions” that pair sound-interfaces withmeaning-interfaces (sounds with meanings/concepts), and presumably thesefeatures make a difference in how an expression’s derivation proceeds and/or inwhat appears at the relevant interfaces. On this assumption, Chomsky (1995c)suggests that lexical features include phonological (sound-specifying), formal(N, V . . .), and “semantic” (artifact ) ones.8 A set of sets of these three kindsof features is a “lexical item.”9 A list of phonological, formal, and “semantic”feature clusters is a list of the features that might compose a lexical item. Atheory of what the features are, of how they can be arrayed and arranged, andhow they can enter and play a role in a linguistic derivation, is a theory of thelexicon and of the role of lexical features in a linguistic derivation.

Thinking of a theory of the lexicon in this way, we can ask how one wouldknow that a hypothesis about a specific word’s features is accurate and ade-quate. Evidence about a specific lexical item’s features in a specific I-languagewould be needed, of course. That might come, in part, from observing how aperson uses a word. But to be sure that we have a good theoretical description ofthe features of even a single lexical item in a single I-language, we need assur-ance that the list of features and possible configurations appealed to in givingthe description is adequate. That is, one needs a list that is descriptively adequate(yields good theoretical descriptions of lexical items’ features) and complete,so that it can – so far as natural languages and their acquisition are concerned –adequately distinguish any specific lexical feature from others. Without this,there is no guarantee that the theory has the right theoretical terms. Furthermore,one needs to consider the role of a lexical item in linguistic processing. If theissue is which features are relevant to defining not a sound, but a meaning – aconcept as it appears in a sentence at SEM – the theory will have to say whichfeatures can appear at the semantic interface of the language faculty, and howthese features “interact” with the features of other lexical items, not just to formphrases and clauses but in more complex ways (consider fake owl and longtrip). That is because it is at that interface – SEM – that the language facultymakes its contribution to a person’s cognitive efforts, to a person’s attempts toclassify, describe, explain, speculate, soliloquize, conjecture, remonstrate, etc. –in effect, to “express thoughts.” Obviously, people use linguistic concepts toperform these tasks; if so, a theory of language needs to say what these con-cepts at an interface “contain” that makes it possible for language to contribute –often remarkably subtly – to the performance of these and other cognitive tasks.Furthermore, it is reasonable to suppose that the language faculty is sufficientlywell “engineered” that at the end of the derivation of a sentence, everything thatappears at SEM be able in principle to “instruct” other systems – in effect, afeature that appears there must be able to make a difference to (or be “legible”to [Chomsky 1986, 2000a, b]) those other systems. So a theory of the languagefaculty and the lexicon should speak to this by ensuring that nothing appears


212 James McGilvray

in lexical specifications that cannot contribute either to the ways SEMs can beproduced (derived) or to “communication” with other systems in the head. But,if so, to be assured that one’s list of lexical features is adequate, one must alsohave in hand an adequate theory of what any lexical item can contribute – forany actual or possible I-language – to SEMs. The theory of the lexicon cannotbe just a theory of one of Harriet’s I-languages, much less one of Harriet’s“words.” It must be a theory that can deal with the lexicons of any I-language,for any natural language (Swahili, Miskito, etc.) either in existence or biolog-ically possible. To have a theory adequate to a specific lexical item is to haveone that is adequate to any; it is to have, at the least, a theory of UniversalGrammar, or UG. This gives an idea of why it has been difficult to develop afull theoretical vocabulary for lexical concepts. No wonder Webster’s authorsdo not define words.

Because humans acquire and use language so readily, it is tempting to thinkthat a theory of meaning should be easy – that it is enough to sit down andruminate, exercise intuitions about how people use language, talk to friendsand colleagues, perhaps read a few articles. Several philosophers have donejust these things; and some have made some interesting observations concerninghow people use words. But their accomplishment is little different from that ofthe dictionary scribe. It is useful, sometimes, for some purposes. The importantissue is not describing usage and exercising the conceptual powers (“intuitions”)of people who have a language, but explaining and describing articulately, ina science, what a language is and what it is about it that lets us use it in thecreative ways we so obviously can. That is work in a natural science and –unaided by rich native conceptual resources – it is very much more difficult.

“But it’s just syntax!”

UG when complete can fully describe all the expressions – including theirmeanings – that are available in an arbitrary I-language. It will offer a scienceof linguistic meaning. It will allow one to define anything that appears at anypossible SEM. But – someone with lingering hopes of a theory that speaks tolanguage–world relations might say – UG is a (broadly) syntactic theory, anda syntactic theory can’t offer a theory of meaning! It only provides descrip-tions of syntactic entities,10 such as abstract versions of [S[DP[DThe] [Nhorse]][VP[Vkicked] [DP[Dthe] [Nstall]]] – versions that do not even have “words” inthem, only theoretical feature-descriptions. How can these be descriptions ofmeanings? Chomsky grants that SEMs are syntactic entities, of course. That ishow they are defined. But they are still theoretical descriptions of meanings.Meanings are syntactically defined – but they do just what linguistic meaningsshould.

A description of what is to be found at SEM is a description of a very richcognitive resource. As indicated, Chomsky includes “semantic” features among



the lexical ones. After derivation, features such as animate , abstract , andartifact “appear” at the (a) interface(s). They are features that, when clus-tered together in a set along with relevant formal features, are sufficient to dis-tinguish one lexical item’s specific linguistic conceptual content – its cognitivecontribution at SEM to other systems – from another’s. Because Chomsky’sterm “semantic” for these features suggests to many ears relations to theworld, to avoid misunderstanding I will call them “FG-features” (“fine-grainedfeatures”). FG-features do not distinguish the noun bank from the two forms ofverb bank (transitive and intransitive). Categorial features (N, Vtr, Vit) do that.Instead, FG-features capture theory-relevant differences between each instanceof bank in each category. There are no settled terms for these differences, so Iinvent enough to make the point. The group of noun banks that can be used torefer to a place for holding items of some sort (financial bank, poker bank, jobbank, etc.) can be thought of as having repository FG-features that distin-guish members of that group from the members of the berm bank group (thenouns that can be used to refer to sides of rivers, to mounds, to billiard tableedges, to cloud configurations, etc.). Both classes of lexical items may also bedistinct from the tilt bank class – the skew of the airplane as it makes a turn.Probably further terms will be required to distinguish different kinds of repos-itory banks (financial, job, computer memory, etc.), but the point is clear. Inhaving a lexical item, a person has remarkably detailed and specific linguisticknowledge. If so, and if FG features appear at SEM, and if SEMs contributeto other systems, it should not be hard to begin to see SEMs in a new light,to see them as “conceptual,” even though they remain syntactic. In effect, theyrepresent (contain) detailed linguistic knowledge, “information” that appearsat SEM. Chomsky mentions many examples without – understandably – beingdefinite about how to represent lexical detail.

Lexical features seem to be adapted to human cognitive purposes – biolog-ical (including social, project-related intentional, etc.) interests and needs, notto the needs of logical consistency. Many nouns have both abstract andconcrete features (Chomsky 1977a, 1995c, 2000a). If something is con-ceived as a village (something is described using the concept villaage ) thatsame thing can be conceived as destroyed by a flood, but also as moved, evenafter a specific concrete instantiation of it has been destroyed. That is becausevillage has the options abstract or concrete , or both.11 So one cansay, “Riverdale will be flooded and destroyed in an hour; we’d better rebuild itup the hill.” If village is also a FG-sub-classification of polity , perhapsthe same is true of all concepts that have polity : town , city , state , etc.Similarly for some other Ns: if something is a door of wood, it is rigid; butone can also go through the door. A book weighs a kilo and is a compellingstory.

When we use sentences with concepts (semantic features), we show that weare conceptually sensitive to the features that their words have: even if we do not


214 James McGilvray

articulate the differences, they appear to configure our assumptions, presuppo-sitions, beliefs, and expectations. For example, many words are (misleadingly)called “container words” (+affected on the outside words?). To betold that someone has painted a house is to assume – unless there is indicationto the contrary – that its outside has been painted. To be close to the outside wallof a house is to be near the house, while to be close to an inside wall of a houseis not to be near a house. If in a cave and looking at its side, one is looking ata side of the cave, not a side of the mountain it is in. To look outside the caveat a mirror that reflects the outer surface of the mountain is, however, to lookat the mountain’s side. Notice in these and other cases that human interests,tasks, and intentions are somehow reflected in the FG-features that distinguishlexical items. This is particularly obvious with words for artifacts – refrigerator,computer, table – reflecting distinct human uses, serving human interests. Thedistinctions can be subtle. To go to a couple’s house is not necessarily to goto their home (this distinction is not always “lexicalized” in other languages).Human concerns and intentions play an important role in our views of the iden-tities of things. To burn a house is to destroy it. To dismantle a house, take thepieces somewhere else, and rebuild is to rebuild that house. To replace parts of ahouse (bricks, boards, beams) one-by-one and take the replaced parts elsewhereand build a house is not to rebuild the original house. Words for natural objectsrespect human intentions and actions too. To put a teabag in water makes thewater into tea. If one’s water supply is filtered at the water works through afilter made of tea, what comes out of the tap is water, not tea. There are innu-merable examples of these and other sorts. So virtually any natural languagelexical item (not scientific term) is a rich source of fine-grained distinctionsthat can be used by a person because they are a part of that person’s linguisticknowledge, represented in lexical features. And these distinctions are providedat SEM to “other cognitive systems” – perhaps vision, perhaps something more“central.” Ultimately, they appear as differences in naturally constituted cog-nitive tools that people can use to – among other uses – describe the worldcreatively.

The richness and detail of specific lexical items is greatly enhanced andrefined when several are put together to compose a phrase or sentence. In sen-tences, but not lexical items, themes (doers of deeds, recipients of goods, etc.)are assigned, tenses specified, scope and specificity indicated, “agreements”fixed, etc. Ambiguities can arise: They are flying planes. Detail and focus ofmany possible sorts become possible: the dog that bit the cat that ate the ratthat . . . Co-reference comes to be specified (or not): Harriet washed herself;Harriet washed her. Mood (interrogative, declarative, etc.) is specified. Inten-sification, diminution, and the like, become possible: very, very, very tediousspeech. There is something like coordination of features: long trip, long walk,long conversation, long paper, long drink, long speech. More room is provided



for imagination and speculation: the amiable person, amiable trip, amiable dog,amiable grouch, amiable table(?!). A potential for metaphor and other figuresof speech arises: Tom the wolf, pussycat, suit, etc. (Unless used and “appliedmetaphorically,” note, these are not metaphors, just complex meanings.) Phrasesand sentential expressions provide at SEM extremely rich and detailed “perspec-tives” (Chomsky 1995b, 2000a), composed and focused in ways that throughmultiple features in multiple configurations support any number of uses. Theseare extraordinary cognitive tools. They do what meanings should.

Again, linguists are nowhere near providing a full list of the set of distin-guishing features, nor are they certain about the kinds of configurations intowhich these features can go. They are not even sure whether they may haveto postulate as many feature-terms as there are lexical roots. Will a relativelysmall set of FG-features and their possible configurations be sufficient to distin-guish all lexical items from each other, or will the lexicologist have to providehundreds of thousands, perhaps millions, of lexical root terms to capture thefinest distinctions between not just all active lexical items in all people’s mentaldictionaries, but all the possible ones? Are the different roots for the noun andverb wash and the noun and verb rinse to be thought of as unique and indecom-posable (rinse and wash), or decomposable instead (rinse = F1 . . . Fn)?This raises the acquisition issue mentioned before: making sense of hundredsof thousands of latent concepts looks more difficult than dealing with a smallerset of features that can be put in arrays. There is also a modularity issue: dog ,for example, has both visual and linguistic features. Where does the linguist’sresponsibility end, and the visual scientist’s begin? Do visual features entera linguistic derivation? Do they come to be coupled with linguistic features“after” a derivation? A full theory of UG would speak to this issue too, if onlyby saying what language’s contribution is. In spite of the unanswered ques-tions, though, there seem to be FG features, and they seem to capture part ofthe conceptual “content” the language faculty contributes to cognition. In thisrespect, they can be thought to capture the linguistic part of what we think of as“understanding.” That qualifies them, surely, as capturing linguistic meaningsor linguistic “contents.”

Perhaps by now FG-features – specifically, their detail and capacity to servebiological (interface) and human interests too – make SEMs (which as mean-ing interfaces are already cut out to play a conceptual role) look more con-ceptual and less syntactic. They can be made to appear even more conceptualby emphasizing that SEMs are mental events that can be seen as constitutiveof human experience and understanding. Emphasizing that SEM contributes toother systems might suggest that, but a lot more can be said about language andits place in mind. Some details appear below. (For more, see Chomsky 1966:65f/2002a: 98f; 1975, 1980, 1995b, 2000a; McGilvray 1998, 1999: chs. 2 and 6,2002a, b).


216 James McGilvray

Meaning in syntax and its use

Syntax broadly conceived is essentially an internalist discipline and, when aform of scientific theory, it is a naturalistic discipline. Natural language syntaxis the study of the intrinsic, internal features and operations of a language.FG-features are syntactic; in essence, they are the “semantic” features thatChomsky suggests are carried along in the derivation of an expression and thatmake a difference at SEM. They – and phonological and formal features –are internal and intrinsic features, defined within a naturalistic theory – that is,intensionally defined.12 This broad view of linguistic syntax includes the subjectmatters of morphology, phonology, lexicology, and – of course – syntax asusually conceived, which deals with categorial features (N, V, etc.), “functional”features such as T(ense) and AGR(eement), and issues such as how lexical itemsthought of as sets of features come to be put together (Merge), how and whysome features move (Attract/Move), and how some come to be deleted (Erase).Syntax as usually conceived – the study of Merge, Attract/Move, etc. – can becalled “narrow syntax.” Chomsky’s theory of meaning, like his theory of sound(phonology/phonetics) is syntactic, but broadly syntactic.

Syntax need not be linguistic. The internalist study of other mental systems,and of relations between the language faculty and those systems, constitute aneven broader notion of syntax. Syntax so conceived is the internalist study of allinternal operations of the mind – any conducive to theory. Internal operationsmight include relations between language and vision; they certainly includerelations between core language faculty operations (CHL) and performancesystems such as those involved in perception and articulation. This very broadsyntax also includes important parts of what has traditionally been thought ofas formal semantics. Chomsky remarks this in 1999 when he speaks of syntaxas including

the study of [internal, mental] symbolic systems, including whatever computational/representational system we take to be internal to the mind/brain. That includes formalrelations among the elements of these systems (e.g. rhyme and entailment, insofar asthese are formal relations among internal symbolic objects), model-theoretic semantics(insofar as the models are considered to be internal objects, i.e. “mental models” – as inpractice they are, in my opinion, contrary to what is often asserted), formal semanticsbased on a relation R (sometimes called “reference”) holding between symbolic objects(e.g. between “London” and its “semantic value,” not an entity in the world, or even theworld as we conceive it to be, but of some internal system of thought that is [perhaps]itself related to the world), etc. (Transcript of speech)

Thus, relations between SEMs and a “commitment box” might also be placedin this very broad form of syntax, as – for example – the formal semantics thatPietroski and Crain appeal to in order to express (the guidance of) meanings.Whether there is a syntactic science of these internal objects and relations is



an open question that depends – among many other things – on whether onecan find the relevant objects and relations and construct a naturalistic science ofthem. There is reason to think there might be one. Perhaps what philosopherscall “analytic truths” are crude descriptions of such relations. As Pietroski andCrain suggest, if Joan uses Harriet persuaded Harold to walk the dog and holdsthe claim to be true, she is also committed to holding that Harold intended towalk the dog. Plausibly, that commitment is due, in part, to the fact that this formof persuade has built into it the feature cause that structures the speaker’sperspective so that – for this case – Harriet is construed as causing a state ofHarold, specifically, his intending to walk the dog. So far, this is all internal tothe perspective and the language faculty itself. But if it is seen as constraining aspeaker’s commitments, it has consequences for what happens beyond the core,combinatorial operations of the language faculty and its semantic interfaces.Perhaps these constraints can be thought of as “instructions” (Chomsky 1995c)that a SEM gives to “other systems” – in this case, a person’s “commitmentbox.” Perhaps they “say,” “If you hold that p is true, you are also committed toq.” But as the tentative terminology suggests, it is not clear what the systemson the other side of SEM are, nor how they are “instructed.”

The success or failure of efforts to construct sciences of relations to systemson the other side of the linguistic interfaces is independent of the issue of whetherthere can be a broad syntactic core science of the interfaces – meaning andsound – themselves. There already are such theories; there is, then, a syntactic,albeit incomplete, “theory of meaning” – an internalist account of understandingby means of language. This theory is in no way a theory of language use. Butit does say what linguistic conceptual resources are available for people to use,and – given the nature of the language faculty that is revealed by Chomsky’stheory of language – it assures us that what is provided to other systems is usable(“interpretable,” “legible”) by them. Chomsky (2000a) speaks of meanings as“guiding” their use. Descartes might have spoken of a will’s freedom to judgebeing constrained by some “ideas” of the “understanding,” particularly if theseideas are “clear and distinct.” Perhaps the discussion above of persuadeindicates what guidance and constraint consist in and it is possible to constructa science of constraint along the suggested lines. But it would still be a syntactic,internalist theory.

Perhaps – as Chomsky sometimes tongue-in-cheek suggests – a Martian orsome other unworldly creature with a different form of mind and understandingthan ours will be able to construct a science of language use. Perhaps our Martianfriend will tell us about it – in a Transgalacticese idiom we can translate intoHumanese or Humanscientificese. Until then, we had better settle for an effortto construct a complete theory of linguistic meaning on internalist and nativistgrounds. That – and forays into other internal systems – should keep us busyfor a very long time.


218 James McGilvray

Linguistic tools that enable creativity

There are interesting connections between Chomsky’s views and LudwigWittgenstein’s. Wittgenstein told us to think of the words of languages as tools.He also suggested thinking of the meanings of these tools as the functionsthey have in “language games” (ways we use language). Chomsky agrees withWittgenstein that if meanings are uses/functions or applications of words (theirrole in truth-telling, to mention one popular option), there is no theory of mean-ing but only commonsense descriptions of use. Putting this in a Chomskyan–Cartesian way, words’ functions are the uses to which they are put by freeagents, uses that might be “incited and inclined” by circumstance (Chomsky1996a, b) but are certainly not determined by them, whether internally or exter-nally. Chomsky differs from Wittgenstein in other respects, however. WhereChomsky thinks that words – or at least, the sounds and meanings of words –are natural mental entities that enable us to deal with problems in a flexibleway, not artifacts, Wittgenstein seems to have thought that people manufacturewords and language to meet their needs. Furthermore – and most important –Chomsky (1966, 2002a) thinks that the fact that words are not the products ofhuman effort is precisely what makes it possible for humans to be the creativecreatures they are even at a very early age. Other rationalists made this point,as did Humboldt. Language – or the “ideas” or concepts of language – have tobe innate and part of our nature if we are to manage to so quickly and in such asophisticated manner deal with the world, play social roles, understand others’motivations, and create paintings and poems.

Chomsky’s biological rationalist picture of the mind illuminates this. Onekey is recognizing that language’s meanings are “legible” to other systems inthe head – that language, a modular system, offers its conceptual resources tothe rest of the mind. In recent linguistic work, this is expressed by saying thatlanguage meets “interface conditions” set by the systems with which it com-municates. These conditions state, in effect, that a derivation of an expressionis successful if it yields at SEM a legible or interpretable concept – one thatcan be “useful.” If this is on the right track, one can think of the languagefaculty and its products (SEMs) as virtually designed to be useful to other sys-tems. Of course, we can actually use only those our systems can cope with:extremely long expressions, for example, would overpower them. But this isstill an extremely large number. By one of Steven Pinker’s (following GeorgeMiller’s) very rough estimates based on the assumption that English speakerscan readily understand twenty-word sentences, it would take 100 billion cen-turies for a person just to hear them all (Pinker 1993). Although we are finite, wewill not run out of internally constituted and defined linguistic cognitive tools.And when we add the contributions of other systems, and the opportunity that aflexible form of cognitive organization affords, the possibilities seem endless.



We can speculate about what the systems with which language communi-cates might be. Plausibly, they are any of the systems that help us carry out thevarious projects that we humans engage in – the tasks in which language andits distinctive “perspectives” so often play a role. A minimal list would consistof the other known faculties of the mind – core vision and the systems withwhich it is intimately connected, such as facial configuration; object config-uration (Marr’s 3-D representations); perhaps systems that are tuned to bodyconfiguration in both self and others, such as those in which Rizzolatti’s “mirrorneurons” (Rizzolatti et al. 1996) participate; plus – in the case of sign languageusers – systems that “shape” visual scenes with linguistic contours (some of thesame systems that deal with linguistic sounds, Petitto et al. 2000). To this mustbe added audition and the systems to which it is intimately related – music (sys-tems for harmony and dissonance along with associated perceptual systems),phonetic and babbling-recognizing systems, and the like. We can add proprio-ception – which no doubt involves several subsystems – plus touch, taste, and thelike. There seem to be systems sensitive to social/community hierarchy, otherstuned to social “balance” – something like a “fairness” or “justice” system, evi-dent even in small children’s social interactions. There are undoubtedly manyothers. Language relates to all of these, we can assume, in complex ways.13

We use linguistic concepts, and those of our other systems, from time to time,singly (rarely) and in concert, to understand and perhaps solve a massive setof problems that the confluence of our projects with the world and with otherpeople and their projects pose for us. Chomsky calls the overall result of thesecooperative efforts by systems to solve problems “common sense understand-ing” (1975, 1995b, 2000a) – a capacity that children develop early and, thus,a capacity that obviously depends heavily on native endowment. But it is alsoa capacity that is extraordinarily flexible and readily turned not only to crudepractical problem solving, but to art and politics. Unless one is doing science,where commonsense concepts almost inevitably mislead rather than help, werely on common sense for everything from coming to understand the motivesof others and thinking about what we want to do next year to discerning themessages of works of art.

Combining recognition of the fact that language produces an in-principleindefinite number of highly specific, distinct perspectives, that vision is config-urable to an indefinite degree (although we seem natively equipped to recognizeand analyze only in certain specific ways), and that language and other systemsrelate to each other in a flexible manner, the conclusion is inescapable. Thehuman mind is virtually designed to be flexible – almost as if it were a general-purpose cognitive device, although it clearly is not. When, furthermore, we addto this that language can be spontaneously internally stimulated – it does notdepend heavily on “input,” as vision typically does – and that other facultiescan too, at least to an extent (as imagination reveals), we find that the human


220 James McGilvray

cognitive system is not only flexible, but can detach itself from circumstanceand range widely. We can speculate, engage in wonder, and cultivate what Kantcalled a “free play between the imagination and the understanding.” The humanmind – especially because it is given the particular and unique contribution oflanguage – is virtually designed to create, and does so, even at a very early age:children quickly engage in fantasy and play. To summarize: that we have innateconcepts that do not have to be learned, that these concepts can be combined inendless fashion in sentential meanings, that they can be readily produced, andthat they relate to other systems in flexible ways, makes possible creativity andthat form of freedom that gives us a distinctively human form of satisfaction.We have these systems and their arrangement and enjoy using them.14

Science and scientific concepts are a different story. They come late, aremanufactured in creating theories (are artifacts), and the project of science (andhigher mathematics) seems to require maintaining a degree of regimentation inthe use of scientific and mathematical symbols. Gottlob Frege’s semantics formathematics depended on this regimentation in use; since his time, “semantictheorists” have tried to apply Frege’s semantics for mathematics to everydaylanguage use. Wittgenstein warned against this. Chomsky heeds the warning –and honors the fact of creativity.

Creativity in the use of language

We are not all poets or innovative scientists, but we all have what might be called“ordinary creativity” (Chomsky 1966, 2002a). Descartes and some of the earlyCartesians discussed the facts of this sort of creativity and their implications forthe sciences of language and mind. The facts have not changed, and neither –Chomsky thinks – have their implications. First noted in print by Descartesin 163715 and worked out in more detail by Descartes’s follower Cordemoy,later rationalists respected the creativity observations and the Romantics (onlya few of whom – A. W. Schlegel, Wilhelm von Humboldt, and Coleridge –were also nativists) developed them into a major theme. Chomsky summarizesthe various aspects of these observations in many writings. Language use “istypically innovative, guided but not determined by internal state and exter-nal conditions, appropriate to circumstances but uncaused, eliciting thoughtsthat the hearer might have expressed the same way” (1996a, b: 2). “Innova-tive” is sometimes glossed instead as unbounded, “uncaused” as stimulus free;appropriate is sometimes paired with “coherent.” The phenomena these termsdescribe are readily apparent to anyone willing to spend a few minutes observ-ing their and others’ use of language. First, ordinary language use is not causedby “input” (it does not correlate with external or internal circumstance andso seems stimulus free); anyone might say anything, anytime, without regardto circumstances. Second, use is typically innovative, novel, or unbounded: a



few formulae (“Hello, how are you?”) aside, repeated sentences are the excep-tion, not the rule. There does not seem to be any upper bound on what peoplecan produce. Nevertheless, third, language use is appropriate and coherent tocircumstances, whether those of the speaker’s present perceptual environmentor circumstances elsewhere and elsewhen, including those of fictive worlds(circumstances in some world speculated about, conjectured, constructed infiction, imagined, hypothesized, etc.). Unbounded, stimulus free, and appropri-ate are all aspects of what Chomsky has in mind by pointing out that ordinarylanguage use is “guided but not determined by internal state and external con-ditions.”

Imagine Gertrude at a party. Her boyfriend is discussing with equally enthu-siastic friends the comparative speeds of execution of the latest computer chips.Gertrude says, “I’m going to join the Canadian bobsled team.” Her environmentdoes not cause this sentence. She need not say anything at all, and could havesaid any number of other things. But the sentence (and an unbounded number ofothers) is “appropriate to circumstances.” Perhaps she is letting her companionsknow that she is bored and wants to talk about something else, or remindingthem that their meals are getting cold. Perhaps she really wants to join a bob-sled team. So while circumstances do not cause her sentence, it is appropriate tothem: she has a reason – perhaps several – to say what she does. Her companionsmight be surprised. But they, and we, have no difficulty explaining (in termsof reasons) why she says this, nor in adjusting with extraordinary sensitivity tosubtly different circumstances, or any number of other sentences.

There is no prospect that the phenomena of creativity can be the subjectmatter of a science of language use, or a science of mind. The science of minddoes, however, say something about the phenomena of creativity. Chomskyremarks (1972a: 13) that while no science can “explain intelligent behavior,”it might explain how intelligent behavior is possible. The creative aspect oflanguage-use observations focus on paradigm cases of intelligent behavior –appropriate but uncaused judgments. A good science of the mind should showhow this form of intelligent behavior is possible. The picture of the mind drawnabove shows how. I summarize to conclude.

Making creativity possible

Think of a scientifically based picture of the mind as presenting the machin-ery of the mind (as understood by the best relevant sciences) as having certainintrinsic features and standing in relationships to each other in ways that makesense of each module’s contribution to human cognition and action. For lan-guage, the features to be made sense of are the unboundedness, stimulus free-dom, and appropriateness of linguistic action. Unboundedness is made senseof by a language faculty that can in principle produce unlimited numbers of


222 James McGilvray

expressions. As for stimulus freedom, it is connected to a cluster of phenom-ena connected with modularity – to the fact that language in particular doesnot require external stimulation, to a degree of autonomy in the operations ofspecific faculties, and to flexible interrelationships between modules. Appro-priateness is the most interesting. Plausibly, linguistic appropriateness can arisein minds that have language, plus a flexible form of overall organization, anda need/urge to “solve problems” of all sorts – practical, intellectual, artistic,etc. To have a concept is to have something like a focus that people can use toarray, categorize, evaluate, etc. In performing a cognitive task, seeking a solu-tion to a cognitive problem, interpreting others, constructing and understandingmetaphors, etc., we routinely distinguish relevant applications of concepts fromirrelevant – appropriate from inappropriate. The endless conceptual resourcesand built-in flexibility of the human mind make sense of how this is possible.They also make sense of why there will never be a specific class of expressionsthat is appropriate to a specific circumstance. As our pervasive use of metaphor,irony, and other figures of speech suggest, our minds are not only tolerant ofinnovative ways of understanding, but we find novel yet apt (fitting, plausible,interesting, intriguing, etc.) ways of understanding a situation satisfying.

It is remarkable that everyone routinely uses language creatively, and gets sat-isfaction from doing so. It is remarkable because unless one is willing to appealto something like Descartes’s second substance, or to theological “explana-tions” that hold that humans are products of a special creation that gives us thisgift, the fact of creativity must somehow be lodged in our natures as biologicalcreatures. A Chomskyan picture of a human mind that has flexibility biolog-ically built into it in the form of endless (although biologically constrained)conceptual resources and minimal constraints on cooperative use of multiplefaculties is a mind with free will. That mind can be turned to the performanceof endless numbers of tasks. It can also be freed from current practical concernsto speculate and imagine.

So humans have the opportunity for linguistic creativity – and for all that thisgives us, given the way language contributes to overall functioning – built intotheir biological natures. That helps explain why Chomsky seems to think thatone of our fundamental needs as humans is to exercise this opportunity. Ful-filling our natures gives us satisfaction. If so, it is not surprising that Chomskyassumes that an ideal form of social organization must provide ample opportu-nity for people to do so.


11 Market values and libertarian socialist values

Milan Rai

Many observers have been fascinated by the puzzle of tracing connectionsbetween Noam Chomsky’s extraordinary contributions to two quite distinctdomains of modern culture. Ronald Lunsford points out that some of these par-allels would exist whatever area of inquiry Chomsky had chosen for his life’swork – parallels that “arise from Chomsky’s way of thinking, rather than fromthe disciplines themselves” (Haley & Lunsford 1994: 172). One critical elementin his “way of thinking” manifest in both his professional and non-professionalwork is his commitment to rationality (a commitment summed up in the obser-vation, “There are no arguments that I know of for irrationality” [Chomsky1987a:22]). On a more abstract level, one finds in both his political and his non-political work a quality that Chomsky has sometimes referred to as “psychicdistance.” In linguistics, “if we can establish a kind of psychic distance fromthe object and try to see how similar normal common characteristics really are,against the background of a possible variety that can be imagined,” we discoverthat “language structures really are uniform” (1988a: 151f). In politics, oncewe have extricated ourselves from conventional thought, we discover under-lying similarities between apparent polarities – an underlying commitment toUS power shared by, for example, Joseph Alsop (a Vietnam War “hawk”), andArthur Schlesinger (a Kennedy liberal and a “dove”).1 The ability to achieve“psychic distance” is actually an essential element of the ability to inquire, andto achieve scientific progress.

Chomsky himself has noted some substantive (though tenuous) connectionsbetween his work in linguistics and his political writings, centred on his con-troversial conception of human nature. He notes that for the last few centuriesof Western scientific thought it has been assumed that “physical structuresare genetically inherited and intellectual structures are learned” (1988a: 399).In his view, on the other hand, a view that he traces back to Enlightenmentthinkers, the mind is composed of many numbers of genetically-inherited sys-tems, “each intricate and highly specialized, with interactions of a kind thatare largely fixed by our biological endowment” (1988a: 500). Some of thesemental “modules” or “organs” are concerned with language, the organization

225


226 Milan Rai

of perceptual space, a theory of number, and so on. In a different region, con-cerned with social relations, there may be a mental “module” devoted to therecognition of personality structure, “which undoubtedly is also a complexand creative intellectual achievement” (1988a: 240). Chomsky suggests that“among the biological characteristics that determine the nature of the humanorganism” there are “some that relate to intellectual development, some thatrelate to moral development, some that relate to development as a member ofhuman society, some that relate to aesthetic development.” Chomsky expectsto find that to a large extent, these characteristics are “immutable”: “That is tosay, they are just part of being human the same way as having legs and arms ispart of being human” (1988a: 147).

To be clear: just as it is the capacity to develop, or “grow,” a human languagethat is innate, and not any particular language itself, so it is the capacity todevelop a human moral system that is innate, not any particular morality. (Sim-ilarly for the capacities to relate to other humans in society, and to appreciateand create aesthetic experiences such as myth and ritual.) The moral “faculty”is not a universal moral code hidden in our DNA, but a mechanism for buildinghuman codes of morality. Chomsky notes that we do develop “implicit systemsof moral evaluation” which are more or less uniform from person to person, withonly very limited exposure to the moral standards of a particular society. Whilethere are certainly moral differences, “over quite a substantial range we tend tomake comparable judgements, and we do it, it would appear, in quite intricateand delicate ways involving new cases and agreement often about new cases,and so on, and we do this on the basis of very limited environmental contextavailable to us.”2 The standard “poverty-of-stimulus” argument applies. Thereis, Chomsky argues, something in human nature that is common across differentmoral systems. Even at the outer limits, we still find Nazi leaders proclaimingthat their actions were in pursuit of the common good. Chomsky notes that itis a rare person who will straightforwardly admit entirely selfish and ignoblemotivations. “Even Himmler didn’t say that” (Chomsky 1993c: 65). (In pass-ing, we may note Chomsky’s suggestion that part of the human “moral faculty”is the genetically inherited need to protect our own self-image – this can giverise to racism, a belief formation needed by oppressors to maintain a positiveself-image [1993c: 72f].)

The notion that there are structures of “unconscious knowledge” within themind, “principles which form the sinews and connections of thought, but whichmay not be conscious principles, which we know must be functioning althoughwe may not be able to introspect into them,” is an insight of the seventeenthcentury expressed in modern language.3 Chomsky has reasserted the place oflinguistics within a modernized form of this rationalist conception of the humanmind. This modernized rationalism can be seen as the intellectual hinge betweenChomsky’s work in linguistics and his political thought. Being careful to


Market values and libertarian socialist values 227

distinguish between rationalism in this sense, and a commitment to rationality,4

we are in a position to unfold some of Chomsky’s thoughts on moral values.

The social consequences of human nature

Chomsky argues that notions of human nature are fundamental to politicalthought and action. He suggests that every kind of political intervention is actu-ally based on a conception of human nature, a speculation or belief concerningwhat is the essence of “being human.” This is so, Chomsky argues, whetherone is proposing social reform, the status quo, reversion to a previous era, orsocial revolution – “any vision that one has as to the nature of a better or utopiansociety, a society towards which we ought to strive, or (to be more incrementalabout it) any point of view that one takes towards the next small change in socialevolution” (1988a: 245). Chomsky suggests that “if there is any moral characterto what we advocate, it is because we believe or are hoping that this changewe are proposing is better for humans because of the way humans are.” Forthe change to be justifiable, there must be “something about the way humansfundamentally are, about their fundamental nature, which requires that thischange we are advocating takes place” (1988a: 597). “Thus, if we are opposedto slavery, it is because we think that in some sense these institutions are aninfringement on essential human nature” (1988a: 385), “because we think thatslavery is inconsistent with fundamental human rights, which are rooted in thenature of humans, which demand that they be free and not owned by others,”because it is part of “essential human nature, an essential human right to be freeand under one’s own control” (1988a: 597). The same connection to deeperintuitions can be found in the case of, say, a change in tax laws, unless the pro-posal is made purely for opportunistic reasons. If we believe that there is somemoral content to what we are proposing, “if we trace that position to its origins,it will have to do with some assumption about human nature, about the centralhuman nature”: “Otherwise, there is no basis for taking a political position ora stand on issues,” Chomsky suggests (1988a: 598). In his view, while theseconnections are “almost never” opened to inspection, it is “crucially important”to make these assumptions explicit, “and to see whether in fact we can find anyevidence bearing on them” (1988a: 597, 245). Chomsky concedes that it isextremely difficult to make these kinds of assumptions explicit, “and so littleis understood that the structures of the argument and belief that we develop arevery loose” – so loose that “they are largely structures of hope and convictionrather than arguments with evidence.” (“But nevertheless those are the struc-tures that must be there for there to be any moral content to our advocacy andaction” [1988a: 598]).

Within modern Western society, a conventional picture of human nature hasgrown up, with obvious consequences for social form – the human being as a


228 Milan Rai

self-seeking consumption-oriented social atom, “economic man” or “economicwoman.” Chomsky observes that human beings do not really live as maximiz-ers of consumption. It is not true, for example, that in a typical family eachmember exerts herself or himself to secure as much food as possible, at theexpense of other members of the family. “The official values of society arevery remote, I think, from most of our actual life with other people,” Chomskysuggests (1988a: 172). He argues that this concept of the “economic human”is “a psychological absurdity” which leads to “untold suffering” for those whotry to mould themselves to this pattern, “as well as for their victims”: “ ‘Lookout for number one’ is a prescription for demoralization, corruption, and ulti-mately general catastrophe, whatever value it may have had in the early stagesof industrialization” (1981c: 226). Chomsky suggests that it would be “veryliberating” for the wealthy as well as for the poor, for the privileged as well asfor the underprivileged, to be able to live in a society where the human essenceis not defined in terms of maximizing production, and producing “on demand”(1988a: 172).

He asks, “What reason is there to believe the crucial assumption that peoplewill work only for gain in (transmittable) wealth and power, so that societycannot be organized in accordance with the socialist dictum?” (That dictumbeing, of course, “From each according to their ability, to each according totheir need.”) This would be true only if we assume that “applying one’s talentsin interesting and socially useful work is not rewarding in itself, that thereis no intrinsic satisfaction in creative and productive work, suited to one’sabilities, or in helping others (say, one’s family, friends, associates, or simplyfellow members of society)” (1971a: 137). This conventional belief, that withoutmaterial reward human beings would simply vegetate, Chomsky describes asa “degrading and brutal assumption,” and a “strange and demeaning doctrine”(1971a: 139, 143).

Separate from, but related to, this notion of homo economicus, is a pictureof the human essence associated with Adam Smith. Chomsky observes thatif we take the view “characteristic of Adam Smith,” that essential to humannature is the “need to truck and barter,” then we will develop a certain imageof a just and proper society – “an early capitalist society of small traders,”perhaps (1988a: 245). A clear example of the kind of connection between socialadvocacy and presumed human nature described above. Interestingly, despitethe veneration granted to Smith in Western societies, few, if any, of his supportersadvocate this form of society. Furthermore, Chomsky notes, few realize thatSmith argued that under the right conditions, the rich “are led by an invisiblehand to make nearly the same distribution of the necessaries of life, which wouldhave been made, had the earth been divided into equal portions among all itsinhabitants,”5 equality of condition, not merely of opportunity. Smith’s ideashave been comprehensively distorted, Chomsky suggests. When we turn to the



real Adam Smith, we find a stinging critique of European colonialism (Chomsky1993d: 4–16, passim), a denunciation of the human consequences of the divisionof labor (for making laborers “as stupid and ignorant as it is possible for ahuman creature to be”) (1993d: 18), a passionate attack on the “malversion,”“plunder,” “knavery,” “extravagance,” “injustice,” and general “indifference toanything but their own interests” of corporations (joint-stock companies)6 and“the mean rapacity” and “monopolizing spirit” of business elites (“merchantsand manufacturers”).7 Chomsky remarks that Smith’s admiration for individualenterprise was tempered by his contempt for what Smith called “the vile maximof the masters of mankind”: “All for ourselves, and nothing for other people”(1993d: 19).

Really existing free market theory

In contemporary society, we find, on the one hand, rhetorical devotion to the“free market,” and, on the other hand, an equally forceful commitment to vio-lating traditional “free market” theory as developed by Adam Smith and others.Chomsky suggests that, “Neither at home nor abroad does the real world resem-ble the dreamy fantasies now fashionable about history converging to an ideal offree markets and democracy, ‘a future for which America is both the gatekeeperand the model.’ ” A more accurate description of the “New World Order” wouldbe that “the world is to be run by the rich and for the rich”:

The world system is nothing like a classical market; the term “corporate mercantilism”is a closer fit. Governance is increasingly in the hands of huge private institutions andtheir representatives . . . With the rapid growth of TNCs [transnational corporations] toa level at which their foreign sales already exceed all of world trade, these systems ofprivate governance gain undreamed-of power. They have naturally used it to create the“de facto world government” described in the business press, with its own institutions,also insulated from public inspection or influence. (Chomsky 1994:185).

It is widely suggested that the present form of “globalization,” and its humanconsequences, are the result of what “the free market has decided, in its infinitebut mysterious wisdom,” or “the implacable sweep of ‘the market revolution,’ ”or “Reaganesque rugged individualism” (Chomsky 1996a,b: 92f). The reality,Chomsky suggests, is revealed by the fact that the rugged individualists ofthe Reagan administration virtually doubled import restrictions to 23 percent,“more than all postwar administrations combined,” while then-Secretary ofthe Treasury James Baker “proudly proclaimed that Mr Ronald Reagan had‘granted more import relief to US industry than any of his predecessors in morethan half a century,’ ” and David Henderson of the OECD found that during the1980s the United States had the worst record for devising new non-tariff barriersto trade (Chomsky 1994: 107). This is merely one aspect of the reality of the


230 Milan Rai

“market revolution.” A 1992 OECD study concludes that “Oligopolistic com-petition and strategic interaction between firms and governments rather thanthe invisible hand of market forces condition today’s competitive advantageand international division of labor in high-technology industries” (Chomsky1996a,b: 129). The same is true of agriculture, pharmaceuticals, services, andmajor areas of economic activity generally, Chomsky observes. Just as, whendiscussing Adam Smith, one must distinguish between Smith’s ideas and theways in which those ideas have been used for particular purposes, so we mustdistinguish between “free market” theory (much of it based on Smith), and whatChomsky calls “really existing free market theory.” The latter has two elements:restricting government intervention to protect the interests of the general publicin the name of economic efficiency, while at the same time securing considerablegovernment intervention to protect the interests of “merchants and manufactur-ers” – also in the name of economic efficiency: “If you look through the wholehistory of modern economic development, you find that – virtually withoutexception – advocates of “free markets” want them applied to the poor and themiddle-class, but not to themselves” (Chomsky & Barsamian 1998: 17). Cor-porations seek government subsidies to help cover their costs and government“insurance” to protect them from market risks, while keeping profits. “Publicsubsidy and private profit” is the continuing pattern of the “market revolution,”Chomsky argues.

We should note that Adam Smith’s critique of government intervention wasnot motivated by his desire to see corporations allowed free rein, but by his con-cern that the governmental system had become an instrument of the “masters.”“For ‘the contrivers of this whole mercantile system,’ the results of governmentintervention were beneficial, while for the general public there were consid-erable costs; ‘our merchants and manufacturers have been by far the principalarchitects’, and their interests have been ‘most peculiarly attended to’, Smithobserved” (cited in Chomsky 1993d:16). Chomsky observes that in any societycontaining concentrations of private power, one would expect such institutionsto gain power and influence over governmental structures (and indeed over cul-tural institutions, a matter we return to). In the case of the United States, “aselsewhere, foreign policy is designed and implemented by narrow groups whoderive their power from domestic sources – in our form of state capitalism, fromtheir control over the domestic economy, including the militarized state sec-tor.” Chomsky notes that studies of top advisory and decision-making positionsrelating to international affairs in the US government confirm the obvious: suchpositions are “heavily concentrated in the hands of representatives of majorcorporations, banks, investment firms, the few law firms that cater to corpo-rate interests,” and those technocratic intellectuals who serve the same interests(Chomsky 1982: 91). It was only natural, then, that when the US government



began planning for the postwar era during World War II, the State Departmentcollaborated with the business-based Council on Foreign Relations, to deter-mine the “requirement[s] of the United States in a world in which it proposesto hold unquestioned power” (cited in Chomsky 1982: 96). The same objectivewas put clearly by the influential George Kennan while he was head of the StateDepartment Policy Planning Staff immediately after the war:

We have about 50 per cent of the world’s wealth, but only 6.3 per cent of its population . . .In this situation, we cannot fail to be the object of envy and resentment. Our real task inthe coming period is to devise a pattern of relationships which will permit us to maintainthis position of disparity . . . We need not deceive ourselves that we can afford today theluxury of altruism and world-benefaction . . . The day is not far off when we are goingto have to deal in straight power concepts. The less we are then hampered by idealisticslogans, the better.8

This is, of course, a Top Secret document. In public, US leaders and opinion-formers were, and remain, much concerned with “idealistic slogans.” To clarifyKennan’s remarks, we must add Smith’s class analysis: while the interests ofbusiness leaders are no doubt to be “peculiarly attended to” by such a mercan-tilist foreign policy, it is misleading to pretend that the interests of the rest ofthe 6.3 percent of the world’s population are really a priority.

There are, then, two quite separate sets of “market values” in operation today:anti-welfare state, “free market” values imposed on the weak and powerless,both at home and abroad; and the protectionist, mercantilist values held bythe powerful and their associates, committed to preserving their “position ofdisparity” through state intervention. Both kinds of “market values” must bedistinguished from Adam Smith’s ideas of “natural liberty,” which belong toan earlier, pre-capitalist, period. Smith was part of the Scottish Enlightenment,with a genuine concern for the mass of the population. We may contrast hisbelief that one of the “two distinct objects” of “political economy” should be“to provide a plentiful revenue or subsistence to the people, or more properlyto enable them to provide such a revenue for themselves,” with Malthus’sdeclaration that anyone lacking independent wealth “has no claim of right tothe smallest portion of food, and, in fact, has no business to be where he [orshe] is,” apart from what his or her offer of labor will bring in the market.9

Enlightenment values

The Enlightenment was a broad, scattered, phenomenon, but there were certaincommonalities to the thought of the period. Chomsky suggests that the idealsof the Enlightenment were the ideas that “people had natural rights, that theywere fundamentally equal, that it was an infringement of essential human rights


232 Milan Rai

if systems of authority subordinated some to others, the insistence that therewere real bonds of unity and solidarity among people across cultures,” and soon (Chomsky 1988a: 764.) Perhaps the central concern of the Enlightenment,however, was its concern with rationality. The French thinker Condorcet pre-dicted that, “The time will come when the sun will shine on free men [andwomen] who have no master but their reason,” while Kant declared, “Sapereaude, have the courage to know; this is the motto of the Enlightenment” (citedin Outram 1995: 1, 2). Chomsky identifies with such sentiments: “I am a childof the Enlightenment. I think irrational belief is a dangerous phenomenon, andI try consciously to avoid irrational belief” (1988a: 773). Chomsky observesthat, according to a tradition extending from Hume to Bertrand Russell,

Reason is concerned with the choice of the right means to an end that you wish to achieve,taking emotional and moral factors into consideration. Unfortunately, too many moderntechnocrats, who often pose as scientists and scholars, are really divorcing themselvesfrom traditional science and scholarship, and excluding themselves from the company ofreasonable persons in the name of a kind of reason that is perverted beyond recognition.(Chomsky 1975:219)

Chomsky is also a “child of the Enlightenment” in his adoption of the othervalues already described. At this point, it may be useful to distinguish betweentwo elements of Enlightenment thought which Chomsky finds attractive. Onemight be called a “weak” set of commitments – to rationality, to equality,to the unity of humankind, and to honesty and elementary decency. This setof Enlightenment values are all that is needed to subject US foreign policy,or the “market revolution,” or the mass media, to a searching critique. Thenthere is a core of ideas in classical liberalism which we might term “strong”Enlightenment values, concerning freedom and the essence of being human,which flower into a much wider, libertarian socialist, critique of Western society.

Wilhelm von Humboldt, an eighteenth-century libertarian with an interest inboth language and freedom, held that

The true end of man, or that which is prescribed by the eternal and immutable dictatesof reason, and not suggested by vague and transient desires, is the highest and mostharmonious development of his powers to a complete and consistent whole. Freedomis the first and indispensable condition which the possibility of such a developmentpresupposes; but there is besides another essential – intimately connected with freedom,it is true – a variety of situations. (cited in Chomsky 1973b: 177)

The “leading principle” of Humboldt’s thought was “the fullest, richest andmost harmonious development of the potentialities of the individual, the com-munity or the human race.” For Humboldt, “to inquire and to create – theseare the centres around which all human pursuits more or less directly revolve”(Chomsky 1973b:177, 178). He argues that, “freedom is undoubtedly the indis-pensable condition without which even the pursuits most congenial to individual



human nature can never succeed in producing such salutary influences”: “What-ever does not spring from a man’s free choice, or is only the result of instruc-tion and guidance, does not enter into his very being, but remains alien tohis true nature; he does not perform it with truly human energies, but merelywith mechanical exactness.” Chomsky summarizes, adding a final phrase fromHumboldt: “If a man acts in a purely mechanical way, reacting to externaldemands or instruction rather than in ways determined by his own interests andenergies and power, “we may admire what he does, but we despise what he is’ ”(1973b: 179). Chomsky remarks that this is a fundamental critique of the wagesystem, which is in its very nature a system of coercion, whereby persons may“choose” to rent themselves to a firm, under the duress of otherwise “choosing”to suffer deprivation and poverty. The classical liberal principle enunciated byHumboldt – that free, creative, work is the essence of being human – is there-fore an indictment of what has been referred to as “wage slavery,” and classicalliberalism may therefore without distortion be developed into a critique ofcapitalism.

For Humboldt, the principal danger to freedom came from the state as “theinfluence of a private person is liable to diminution and decay, from compe-tition, dissipation of fortune, even death” (Chomsky 1973b: 180). While hecorrectly pointed out that these “contingencies” could not be applied to thestate, Humboldt failed to see that the joint-stock company might one day belegally recognized as a corporate “private person,” with precisely the kind ofpotent immortality that he and other Enlightenment thinkers feared.

Chomsky detects a “line of development” in traditional rationalism, runningfrom Descartes, through the more libertarian Rousseau, through Kantians suchas Humboldt, into the nineteenth century, which holds that “essential featuresof human nature involve a kind of creative urge, a need to control one’s produc-tive, creative labor, to be free from authoritarian intrusions, a kind of instinctfor liberty and creativity, a real human need to be able to work productivelyunder conditions of one’s own choosing and determination in voluntary associa-tion with others” (1988a: 594). In the nineteenth century, this notion of essentialhuman nature was developed into a rejection of capitalism, and the power it con-centrated in the hands of those who owned the productive resources of society,coupled with a rejection of authoritarian state socialism. Liberal ideas evolvedin new circumstances into forms of libertarian socialism. Classical liberalismopposed state intervention in social life “as a consequence of deeper assump-tions about the human need for liberty, diversity and free association.” “On thesame assumptions,” Chomsky argues, “capitalist relations of production, wagelabor, competitiveness, the ideology of ‘possessive individualism’ – all mustbe regarded as fundamentally anti-human”: “Libertarian socialism is properlyto be understood as the inheritor of the liberal ideals of the Enlightenment”(1973b: 157).


234 Milan Rai

Humboldt’s central concern was “the fullest, richest and most harmoniousdevelopment of the potentialities of the individual, the community or the humanrace.” Echoing Humboldt, anarchosyndicalist Rudolf Rocker observed that,“For the anarchist, freedom is not an abstract philosophical concept, but thevital concrete possibility for every human being to bring to full developmentall the powers, capacities and talents with which nature has endowed him [orher], and turn them to social account” (1973b: 151). “In a socialist society, asenvisioned by the authentic left,” Chomsky argues, “a central purpose will bethat the necessary requirements of every member of society be satisfied”:

Individuals will differ in their aspirations, their abilities, and their personal goals. Forsome person, the opportunity to play the piano ten hours a day may be an overwhelmingpersonal need; for another, not. As material circumstances permit, these differentialneeds should be satisfied in a decent society, as in healthy family life. In functioningsocialist societies such as the Israeli kibbutzim, questions of this sort constantly arise.(1987a: 192)

In a decent society, “socially necessary and unpleasant work would be dividedon some egalitarian basis, and beyond that people would have, as an inalienableright, the widest possible opportunity to do work that interests them.” Peoplemight be motivated by “self-respect,” if they do their work to the best of theirability, “or if their work benefits those to whom they are related by bonds offriendship and sympathy and solidarity.” Chomsky notes that such notions are“commonly an object of ridicule – as it was common, in an earlier period, toscoff at the absurd idea that a peasant has the same inalienable rights as a noble-man.” He observes that there have always been, and will always be, those whocannot imagine a social order different from that which exists. “Perhaps theyare right, but, again, one awaits a rational argument” (1973b: 142f). Chomsky’sown “hopes and intuitions” are that “self-fulfilling and creative work” is “a fun-damental human need,” and that “the pleasures of a challenge met, a work welldone, the exercise of skill and craftsmanship,” are “real and significant,” andare “an essential part of a full and meaningful life.” He suggests that “The sameis true of the opportunity to understand and enjoy the achievements of others,which often go beyond what we ourselves can do, and to work constructivelyin cooperation with others” (1988a: 394). “Compassion, solidarity, friendshipare also human needs,” Chomsky points out – “driving needs, no less than thedesire to increase one’s share of commodities or to improve working conditions”(1981c: 224). “Predatory capitalism . . . is incapable of meeting human needsthat can be expressed only in collective terms, and its concept of competitiveman who seeks only to maximize wealth and power, who subjects himself tomarket relationships, to exploitation and external authority, is anti-human andintolerable in the deepest sense” (1973b: 184).



No one who gives a moment’s thought to the problems of contemporary society can fail tobe aware of the social costs of consumption and production, the progressive destructionof the environment, the utter irrationality of the utilization of contemporary technology,the inability of a system based on profit or growth-maximization to deal with needs thatcan only be expressed collectively, and the enormous bias this system imposes towardsmaximization of commodities for personal use in place of the general improvement ofthe quality of life. (1981c: 223)

In modern capitalist democracies, there is a fundamental democratic “deficit”in that the entire world of commerce, industry and finance is excluded fromdemocratic control. And within the capitalist firm, Chomsky suggests, thereare hierarchical structures of authority “of a kind that we would call fascistin the political domain” (1987a: 32). There is an inescapable tension betweencapitalism and democracy, a contradiction which can be resolved either byeliminating democracy, or by extending it to the economy. The economy couldnot be democratized along classical liberal principles through state ownershipand management, but through new democratic economic institutions. Summa-rizing the classical anarchist consensus, Chomsky observes, “one can imaginea network of workers’ councils, and at a higher level, representation acrossthe factories, or across branches of industry, or across crafts, and on to gen-eral assemblies of workers’ councils that can be regional and national andinternational in character” (1981c: 249). Another system might organize alonggeographical lines, with local, regional and international assemblies. Both sys-tems would control themselves directly, without a coercive apparatus abovethem. (Chomsky professes himself agnostic as to how the two systems mightinteract, or whether both are necessary.)

Such a prospect is a rather distant one. Chomsky has warned that any seriousmovement of the left should distinguish clearly between “its long-range revo-lutionary aims,” and “certain more immediate effects it can hope to achieve”(1981c: 222). He has made a point of concentrating on the latter almost to theexclusion of the former. The immediate problems are to help people who areoppressed and under attack, to try to prevent environmental disaster, and tobuild a mass movement of reform. The latter should be “a movement for socialchange with a positive programme that has a broad-based appeal, that encour-ages free and open discussion and offers a wide range of possibilities for workand action” (1981c: 221f).

Chomsky hopes that a movement of reform can nurture more revolutionarytendencies committed to challenging the wage system and other aspects ofeconomic authoritarianism. This more radical movement of the left “has nochance of success, and deserves none, unless it develops an understanding ofcontemporary society and a vision of a future social order that is persuasive to alarge majority of the population.” In the advanced industrial societies, the mass


236 Milan Rai

of the population have far more to lose than their chains – “On the contrary, theyhave a considerable stake in preserving the existing social order.” The culturaland intellectual level of any serious radical movement, therefore, “will have tobe far higher than in the past”: “It will not be able to satisfy itself with a litanyof forms of oppression and injustice. It will have to provide compelling answersto the question of how these evils can be overcome by revolution or large-scalereform.” The left will have to “achieve and maintain a position of honesty andcommitment to libertarian values,” and avoid the temptation to adopt messianicillusions concerning the “vanguard party” (1981c: 222f).

Manufacturing consent

Standing in the way of all these developments, however, is the central prob-lem of freedom of thought. In the case of the United States (and other moderncapitalist democracies) there is considerable internal freedom in principle, butvery little real freedom of thought, Chomsky charges. The dominant picture ofthe mass media is, of course, very different (as is the self-image of the media).Judge Gurfein, deciding in favor of press freedom in the Pentagon Papers case(when the US government attempted to suppress the publication of the leakeddocuments), declared that the United States had “a cantankerous press, an obsti-nate press, a ubiquitous press,” which “must be suffered by those in authorityin order to preserve the even greater values of freedom of expression and theright of the people to know.” Supreme Court Judge Powell, in another case,argued that as “no individual can obtain for himself [or herself] the informationneeded for the intelligent discharge of his [or her] political responsibilities. Thepress performs a crucial function in effecting the societal purpose of the FirstAmendment,” by enabling the public to exert control over the political pro-cess (Lewis 1987, cited in Chomsky 1989: 2). For Chomsky and his co-authorEdward Herman, the “Free Press” serves a rather different “societal purpose”:“It is the societal purpose served by state education as conceived by James Millin the early days of the establishment of this system: to ‘train the minds ofthe people to a virtuous attachment to their government,’ and to the arrange-ments of the social, economic, and political order more generally.” It is thesocietal purpose of “protecting privilege from the threat of public understand-ing and participation” (Chomsky 1989: 13, 14). This is achieved, accordingto Chomsky and Herman, through what they describe as “brainwashing underfreedom.” No central authority determines “the party line” and dictates to theorgans of public expression what can and cannot be said. Intellectual confor-mity is not achieved through violence or the threat of violence, as in the Stalinistsystem of propaganda. The Chomsky/Herman “Propaganda Model” of the USmass media is a “guided free market” model, in which thought is controlledby market forces operating in a highly unequal society. Media institutions are



themselves large profit-seeking corporations, “which are closely interlocked,and have important common interests, with other major corporations, banksand government” (Chomsky & Herman 1988: 14). They are also dependenton advertising revenue from other businesses, who gain a “de facto licensingauthority” (Curran & Seaton 1985: 31) over the mass media. Complementingthese and other institutional factors at the “macro” level is the process of edu-cation, selection, and cooption, by which individual reporters become attunedto the dominant ideology. Recruitment to a mass media organization is predi-cated on the candidate possessing the “right” attitudes. For those few who, oncerecruited, display an unacceptable independence of mind, pressures are soonbrought to bear to help them develop a politically correct set of “news values.”Much the same is true in academia, Chomsky observes:

To put it in the simplest terms, a talented young journalist or a student aiming for ascholarly career can choose to play the game by the rules, with the prospect of advance-ment to a position of prestige and privilege and sometimes even a degree of power; or tochoose an independent path, with the likelihood of a minor post as a police reporter or ina community college, exclusion from major journals, vilification and abuse, or drivinga taxi cab.

Given such choices, “the end result is not very surprising” (Chomsky 1982:14). Another, “supportive,” factor is “elemental patriotism” – “the overwhelm-ing wish to think well of ourselves, our institutions and our leaders,” anotherextension of the “moral faculty” discussed earlier (Chomsky & Herman 1988:305).

Chomsky argues that thought control in free societies is most effectivelyachieved when, rather than setting down a single party line to be followed, theinstitutions of indoctrination set down the boundaries of acceptable opinion,and allow debate to flourish within these confines: “in a properly functioningsystem of propaganda,” debate should not be stilled, “because it has a system-reinforcing character if constrained within proper bounds.” What is essentialis to set the bounds firmly. “Controversy may rage as long as it adheres to thepresuppositions that define the consensus of elites, and it should furthermore beencouraged within these bounds, thus helping to establish these doctrines as thevery condition of thinkable thought while reinforcing the belief that freedomreigns” (Chomsky 1989: 48).

Vietnam

Thus, in the case of the Vietnam War, the official version was that US interven-tion was needed to defend the Republic of South Vietnam against its aggressive,Communist, northern neighbor. When a secret US government history of theVietnam War (the “Pentagon Papers”) was leaked in 1972, those US citizens


238 Milan Rai

who cared to know learned that when Washington determined on its major esca-lation of the war in February 1965, US intelligence knew of no regular NorthVietnamese units in South Vietnam; and five months later, while implementingthe plan to deploy 85,000 troops, there was only concern about the “increasingprobability” of such units in South Vietnam or Laos (Chomsky 1972b: 195).“In the light of these facts,” Chomsky observes, “the discussion of whether theUS was defending South Vietnam from an “armed attack” from the North – theofficial US government position – is ludicrous” (1972b: 1960. Nevertheless, this“ludicrous” assertion, converted into an unquestioned assumption, continues togovern discussion of the Vietnam War in the United States and elsewhere inthe West. Mainstream debate turns on whether it was wise to attempt to defendSouth Vietnam. The notion that the United States did not defend the people ofSouth Vietnam, but rather attacked them, does not figure even as a theoreticalpossibility.

A Pentagon analyst, writing in the Pentagon Papers, noted that South Vietnamwas “essentially the creation of the United States” (cited in Chomsky 1982:376n.27). This “creation” of the United States was the “legitimate local author-ity” which invited US military intervention. Chomsky cites the observationof the London Economist in connection with Afghanistan, “an invader is aninvader unless invited in by a government with a claim to legitimacy.” He com-ments, “outside the world of Newspeak, the client regime established by theUnited States had no more legitimacy than the Afghan regime established bythe USSR” (1987a: 225).

Orwell’s Problem

The power of the US propaganda system is demonstrated by the extraordinaryuniformity of articulate US opinion with regard to Vietnam. This is anotherexample of what Chomsky has termed “Orwell’s Problem.” “Plato’s Problem”is how, given so little evidence, we know so much. “Orwell’s Problem” is “theproblem of explaining how we can know so little, given that we have so muchevidence.” Orwell was concerned with the power of totalitarian systems toinstill beliefs that were firmly held and widely accepted despite being withoutfoundation, “and often plainly at variance with obvious facts about the worldaround us.” Chomsky sums up many years of work, with collaborators andassociates, by suggesting that “thousands of pages of detailed documentation”have demonstrated “beyond any reasonable doubt” that in democratic societiestoo, “the doctrines of the state religion are firmly implanted and widely believed,in utter defiance of plain fact, particularly by the intelligentsia who constructand propagate these doctrines” (1986: xxv, xxvi).

Throughout his career, Chomsky has insisted that “the level of culture thatcan be achieved in the United States is a life-and-death matter for large masses



of suffering humanity” (1971a: 249). The level of honesty and decency in USsociety – the commitment of the general public to the weak form of Enlight-enment values – can restrain US interventionism. The more that the generalpublic pierces the fog of propaganda and comes to understand the realities ofUS foreign policy, the more vocal and widespread is the dissent, and the moreconstrained US policy-makers become, sometimes with enormous benefits forpotential victims in far-off lands. Furthermore, unless and until the generalpublic is able to break free from “brainwashing under freedom,” hopes for arevitalized left-wing movement able to replace present institutions with morehumane alternatives must be rather dim. Therefore, Chomsky urges, “For thosewho stubbornly seek freedom, there can be no more urgent task than to cometo understand the mechanisms and practices of indoctrination” (1987a: 136).Luckily, all that is required for “intellectual self-defence” is “ordinary commonsense,” and “a willingness to look at the facts with an open mind, to put simpleassumptions to the test, and to pursue an argument to its conclusion” (1977b: 5).Chomsky observes that “dispelling the illusions is just part of organizing andacting,” “not something you do in a seminar or in your living room.”10

Moving towards the fulfilment of Humboldt’s classical liberal ideals in lib-ertarian socialist forms, we must strengthen the hold of some rather weakerEnlightenment values (rationality, honesty, decency, and so on) in Westernsociety, exposing and rejecting the use and misuse of “free market” values asideological weapons. It is unclear whether there will ever be any unificationof Chomsky’s two sets of contributions to modern culture in a “science” ofessential human nature, but what is clear is that there is a consistency andintegrity to Chomsky’s work in these different fields. There is an underlyingand unifying respect, as Humboldt puts it, “a sense of the deepest respect forthe inherent dignity of human nature, and for freedom, which alone befits thatdignity” (Chomsky 1973b: 177).


12 The individual, the state, and the corporation

James Wilson

Structures of governance tend to coalesce around domestic power, in the lastfew centuries, economic power.

(Chomsky 1996d: 178)

Introduction

This chapter explores Chomsky’s political morality by focusing on his ideasabout the individual, the corporation, and the state. This presentation is nosubstitute for reading Chomsky himself. His propositions become far moreconvincing upon studying the historical record he has assembled, a dismayingcollection which reveals how the actions and rationalizations of the ruling classare remarkably predictable (at least by the standards of any social science).

I will follow Chomsky’s approach of focusing on the United States to under-stand the interlocking relationships between state and private power. For morethan three decades, he has relentlessly pursued the task of documenting theAmerican political system’s effects on its own citizenry and other nations. Heprovides so many facts to support his claims and verify his predictions thateven the reader who ultimately rejects his analysis is likely to emerge less thanenthralled with most American foreign and domestic policies and more skep-tical of how government officials and the media present these decisions to thepublic. For instance, the idea of an idealistic, naıve United States easily manip-ulated by ruthless, clever foreigners is a sentimental myth. Far closer to thetruth is the fear Simon Bolivar expressed in 1822: “There is at the head of thisgreat continent a very powerful country, very rich, very warlike, and capable ofanything” (cited in Chomsky 1993d: 142; see also Wilson 2002).

Like most social critics, Chomsky has found it easier to criticize than to fullydevelop persuasive alternatives. For many of us, his vision of anarchosyndical-ism or libertarian socialism seems excessively optimistic. But for those of uswho believe that the human race can – indeed must – radically improve its socialpractices, if only to survive, the structure of Chomsky’s critique provides someof the means necessary to develop a humane alternative to the existing order.Imbedded in Chomsky’s political writings are not just numerous expressions of

240


The individual, the state, and the corporation 241

outrage at the injustices of Western societies, but also indications of the valuesand institutions necessary to create a more decent world.

The ideal and the individual

It is a commonplace that we know very little about ourselves. Chomsky puts aparticular gloss on this when he notes that the science of human nature is littleadvanced:

We don’t know anything about human nature. If we’re rational we know that it exists andundoubtedly there are very powerful biological constraints on the way we think, whatwe do, what we conceptualize, what we imagine and our fears and hopes, etc. But aboutwhat they are, you can learn more from a novel than from the sciences. You operate onthe basis of your hopes. (Chomsky & Barsamian 1992: 354)

So while we can assume that our biology defines our nature, and while (contraryto the quotation) we do as scientists know a little about that nature – we havesciences of language and human vision, for example – there is clearly muchmore for scientists to learn about that nature and the constraints or limitationsit puts on what we can think and do. The emphasis is on “constraints” and“limitations” because a science of human nature can speak to what “entersinto” the choices we make and the thoughts we have (it can tell us about ourconceptual tools), but it does not causally explain our particular choices or ourbehavior. It doesn’t tell us how to “operate.” The science of human nature isthus no science of human action; it certainly is not a “behavioral science.” It islimited to determining the processes every person uses to make decisions.

As Chomsky suggests in his discussion of the creative aspect of languageuse (1966, 1988b, 1996a,b, 2002a), science does not have and likely cannotconstruct the tools (concepts) to deal with human choice and creativity. Thatis why crucial aspects of what it is to be a person – an individual’s free willand the specific history of actions that such freedom engenders – are opaque toscientific investigation, now and probably forever. This has a bearing on what itmeans to be an individual. Chomsky, like Descartes, thinks of the person as anintellect associated with a will that freely chooses its course of action: humansare “free agents” who probably have a corresponding “instinct for freedom.”In other words, any political/moral conception of “human nature” cannot bereduced to mathematical or biological models that can accurately predict futureactions.

Another element is reflected in his respect for “the common man” and hisrelated defense of full democracy – a democracy wherein all individuals areempowered to participate actively in all matters of concern to them, includ-ing economic and market choices. Chomsky’s political views are compati-ble with the science of human nature which – even in its present, immature,


242 James Wilson

state – clearly indicates that humans have readily available to them at an earlyage the conceptual materials they need to understand political, economic, andsocial issues and then make political choices. What we know about humanbeings (as we have seen in Chomsky’s solution to the problem of explaininghow children acquire languages so quickly and easily) indicates that everyone –every human, without regard to culture, history, training, background, wealth,etc. – has, or can easily acquire, the conceptual tools needed to understand themotivations and projects of others (whether acting as individuals or groups),assess the value and effects of those projects on their own interests, and makechoices. Everyone has “Cartesian common sense” (1977c). Using this form ofunderstanding, we chat, speculate, describe, soliloquize, wonder, conjecture . . .In gossip and thought, planning and advising, we complain about others, assessthe chances of favorite sport teams, and criticize the policies of governmentand corporation. The shift from this scientific description of a mental power inall individuals to Chomsky’s commitment to democratic principles should beobvious, even if it is a transition from an objective understanding to a norma-tive conclusion. Cartesian common sense provides us all with the tools to dealwith any political issue and, Chomsky holds, everyone ought to be given theopportunity to make decisions about the full range of human affairs.

Still another element in Chomsky’s view of the individual – related to both ofthe factors mentioned above – is the postulate that the use of authority of any sortto control another or reduce their autonomy needs justification. The scientificfact of commonsense conceptual equality supports the normative requirementthat the powerful must provide good reasons for imposing their views on oth-ers. Thus, the exercise of the authority of parents over children, instructorsover students, generals over soldiers, priests over parishioners, corporationsover workers, and states over populations all require justification. Sometimesauthority satisfies that burden of proof. Chomsky points out that it is not diffi-cult to defend a grandparent’s restraining a grandchild from running into a busystreet (Burchill & Chomsky 1998: 13). But in many cases – and especially whendealing with adults where their economic and political interests are involved –it may be very difficult. The principle that authority needs justification is oneof the roots of Chomsky’s anarchism or libertarianism.

A final element in Chomsky’s conception of the individual is his (surelyuncontroversial) assumption that individuals survive and thrive best when theylive and cooperate with others in communities. Individuals need communitiesnot only for support at early stages of development and for fulfilling coopera-tive projects (such as working together to make diesel engines), but also for allthe data that are crucial to cognitive development, including the developmentof language. And the tools needed for cooperation and life in a communityare by virtue of their biology built into everyone. We have the built-in cogni-tive equipment that facilitates love, friendship, sympathy, and support. Peoplehave something like what Hume called a “moral sense,” a capacity not only to



understand the motivations, thoughts, and feelings of others, but also to sympa-thize with and put oneself in another’s place. This is another premise underlyingChomsky’s syndicalism or socialism.

Returning to the quotation, some readers might be puzzled by Chomsky’sassertion that when it comes to comprehending the “very powerful biologicalconstraints on the way we think, what we do, what we conceptualize, what weimagine and our fears and hopes,” we can “learn more [about what they areand how they affect our actions] from a novel than from the sciences.” Betterthan any other art form, the novel mimics the chatter of human consciousnessthrough the author’s voice as well as the characters’ thoughts. Because all ofus have “Cartesian common sense,” we can understand the actions of ourselvesand others by putting ourselves in others’ places. We “get inside” the characters’minds, observing that they feel and react in familiar ways. Novels allow us todo this in a focused, relatively uninvolved way. We appreciate and assess themotivations, decisions, and choices that the novelist’s characters make. Thisfocused exercise of the full range of our native cognitive capacities withoutdirect involvement provides a more fruitful study of “the way we think, whatwe do, what we conceptualize, what we imagine and our fears and hopes” thanthat offered by the findings of linguistics and the theory of vision or the simplisticand skewed models of choice and rationality that permeate contemporary socialsciences. Particularly odious and misleading is the current obsession with asimple and materialistic conception of “rationality” that assumes we alwayschoose to maximize benefits in the form of goods and profits for ourselves.Quite simply, no simple model or theory of choice and motivation illuminatesas much as a good novel.

We must recognize the truth of the Aristotelian insight that in practical affairspeople ultimately rely on a rich form of “practical reason”; they intuitivelychoose between competing sets of arguments that support (but never logically“prove”) different courses of action (Aristotle 1984: 1805). Almost by defini-tion, this conception of intuition is elusive, involving contributions from eachperson’s unique character, predispositions, experiences, perceptions, beliefs,understanding of circumstances, combined with and processed by – somehow –an underlying human nature. Practical decision-making includes all politicalthought and activity and it is fundamentally different than scientific and mathe-matical reasoning. Most of us would agree that humans should treat each otherdecently, but such a statement is nothing like the assertion of a mathematicalprinciple. And when we try to decide about how to be fair in a specific case –how to distribute goods among a group of people, for example – we are unlikelyever to find the sort of consensus that mathematicians achieve when they offerproofs of theorems.

All the above supports Chomsky’s belief that every person has “an instinctfor freedom” (1992b: 335)1 combined with a Humean “moral sense” (Rai 1995:108). By positing such moral premises, Chomsky is proposing his own theory


244 James Wilson

of human operations, even if he is uncertain of their scientific validity. Thusthe libertarian belief in the free and creative agent combines with the recogni-tion of a moral assumption to create a principle resembling Kant’s universalisthumanitarianism: each of us should treat our fellow humans, who by all availableevidence have the same amazing levels of perception, feeling, and understand-ing as ourselves, as equal moral agents – as ends, not as “mere instruments” forpersonal gain and pleasure (Chomsky 1993d: 288n.4).

So how do these premises and distinctions influence particular socialpractices? Unsurprisingly, Chomsky admires the libertarian-romantic Wilhelmvon Humboldt (Chomsky 1993d: 19), who observed “It is the prosecution ofsome single object, and in striving to reach it by the combined application ofhis moral and physical energies, that the true happiness of man, in his full vigorand development, consists” (Humboldt 1993: 3–4). This kind of fulfillment canonly arise when an individual has the opportunity to choose a particular goal andwhen the path to that goal offers a variety of situations: “Even the most free andself-reliant of men is hindered in his development, when set in a monotonoussituation” (Humboldt 1993: 10). Workplace monotony and lack of autonomyare far too often the rule and not the exception. Personal productivity – poten-tially the most important area in which a person can express his or her autonomywhile finding the satisfactions of life in a community – is usually under auto-cratic control. Echoing Adam Smith in the Wealth of Nations, Chomsky warnsof the human costs accompanying the economic benefits of division of laborand “wage slavery” in autocratic structures that minimize or remove workerchoice and participation. Only governmental control and action can eliminatesuch autocratic institutions to guarantee the worker’s full participation in theworkplace (Smith 1076: 303).

Reflecting human beings’ need to exercise choice and to associate,Chomsky’s ideal form of social organization is based on his view of the individ-ual as a creature that needs to be creative and to freely associate with others forthe fulfillment of common aims. Thus, for Chomsky, the ideal form of socialorganization is one that minimizes external authority (anarchism) and allowsfor free association of individuals (syndicalism). The result, which he calls“libertarian socialism” or “anarchosyndicalism,” maximizes the opportunityto exercise autonomy, freedom, and creativity on the one hand, while findingfriendship, solidarity, and love, on the other.

It seems unlikely that anything like Chomsky’s anarchosyndicalist form ofpolitical organization can be fully realized. This criticism differs from that whichmight be offered by a neoliberal who makes (apparently) flawed assumptionsabout human nature. Neoliberals generally posit that humans seek to accu-mulate as many goods as they can and to dominate others (Chomsky 1996b).These features aptly characterize the aims of corporations, but not moral agents,unless they are pathological. The first claim – that people have an intrinsic need



to accumulate as much as possible – is patently false. The second – that allpeople need to dominate others – is dubious, at best. Evidence that peopleseek to dominate, such as that reputedly offered by anthropological study ofmales in the Yanamamo tribe in Central America, or by studies of Igorot tribes-people, does not prove the case: among other things, it focuses on males, andmales during specific phases of growth, not all people at all times. In favor ofChomsky’s universalist position, on the other hand, there is a mass of histori-cal, anthropological, and acquisition data. For example, there is good evidencethat even very young children have an innate conception of fairness and thatthey find self-justification necessary whenever they act unfairly (Burchill &Chomsky 1998: 9). As noted above, Chomsky does not think we know enoughabout human nature to be confident of either polar position. But surely weknow enough to recognize that neoliberalism’s narrow assumptions about ourfundamental needs are not just wrong, but pathological.

The case for some degree of hierarchy and institutionalized authority is basedon five observations. One is that conflicts perpetually arise between humansconcerning ownership, exchange, assault, affront, distribution of goods, andthe like. Experience shows that resolution of these kinds of conflicts is usuallymost effective when it is performed by neutral third parties backed up withthe authority to enforce their solution. People have a hard time reaching fairoutcomes when they are personally involved in a dispute. Third-party disputeresolution requires at least some kind of state-vested hierarchy and, for enforce-ment, a form of authority and power assigned to individuals playing institutionalroles that are constitutionally fixed and vested in a government. Second, man-agement of complex institutions, including workplaces and industries, requiresspecial talents and developed skills that take education, experience, and effortto develop. If so, some form of decision-making authority must be accorded tospecific individuals. For better and for worse, lawyers and judges justify theirauthority in part upon their expertise gained through deep immersion in thelegal system, a process that enables them to “think like a lawyer.” Because bothconflict resolution and specialization of management require vesting author-ity in specific individuals playing specific roles, it is not clear, at least to me,that something like Chomsky’s bottom-up anarchosyndicalist form of socialorganization could be instituted. Third, there is the matter of efficiency: a formof organization in which individuals are randomly rotated through positions islikely to be inefficient. Individuals would have to learn to play their roles andwould likely find that they have become good at it just about the time a newperson takes over. Next, Chomsky’s approach seems internally inconsistent;there needs to be some external authority (which violates any vision of pure anar-chy) to check potential abuses by any anarchosyndicalist institutions. Finally,it is hard to refute neoliberal assumptions completely; history has shown againand again that many people are prone to aggression and greed, just as history


246 James Wilson

contains innumerable instances of compassion. Such a mixed view of humannature prevents me from embracing pure forms of either neoliberalism or anar-chism.

However, I have far fewer worries if one treats Chomsky’s anarchosyndi-calism as what he has called a “vision.” As an ideal, even if not realizable, itserves as a measure to judge whether our current society meets the fundamentalhuman needs of freedom and solidarity that Chomsky’s Enlightenment view ofhuman beings endorses. It can also help us decide what we should do to makeour society approach the ideal.

Chomsky on the corporation and the state: an overview

Humans are, as Aristotle said, “political animals.” Individual human beings arenot self-sufficient, nor are their natures fully satisfied, unless they find recogni-tion, love, solidarity, and friendship from others. In early life they need some sortof family structure to provide nurture, offer a wide range of experience, and –in certain respects – training during a lengthy, vulnerable childhood. Later, theyincreasingly require recognition of their creative activities from others, alongwith the material, social, and individual benefits of organized forms of labor thatallow individuals to develop their specific talents and skills in the productionof goods and services. The forms of organization labor takes can, and over his-tory have, changed. In current, industrialized societies, autocratic corporationsare the predominant form used to supervise labor. Unsuprisingly, Chomsky’shistorically informed critique of the corporation commingles with his view ofthe state.

States have mutated over history. Aristotle’s small city-states, which requiredcitizens to directly participate in legal, defense, welfare, and administrativematters, no longer exist. In current states, there is little except sentiment andpropaganda left of the notion that citizens should fully participate in the govern-ment’s legal, social, and administrative operations. Private corporations play acentral, ever-increasing, role – indirectly through the control of transient capi-tal and directly in the United States through campaign contributions to the twomajor political parties. As a result, a small group of people exercise most mean-ingful political authority. Voting procedures, largely indistinguishable politicalparties, massive campaign funding, corporations given the legal right to influ-ence campaigns, and corporate-run media provide little voice to any but therich. Corporate lobbying plays an extraordinarily important role in influencinglegislation. Transnational corporations and business-influenced transnationalorganizations such as the World Trade Organization (WTO), the InternationalMonetary Fund (IMF) and the World Bank largely control all but the most pow-erful nations. Chomsky the anarchist would prefer to replace all these top-downauthoritarian structures with democratic institutions based on fully participatory



worker organizations (his syndicalism). In circumstances such as those foundin largely corporate-dominated democracies such as the United States andEngland, however, he knows that the sole form of control of authority thatmost people can exercise consists of voting for representatives in local, munici-pal, regional, and national forms of government. Thus, in current circumstances,Chomsky the anarchist paradoxically supports efforts to increase the power ofthe state, at least where it can serve to regulate and check otherwise largelyunconstrained and otherwise unaccountable corporate authority.

Chomsky’s view that in current society we must build up the regulatorypowers of government demonstrates how his ideal of anarchosyndicalism simul-taneously criticizes the status quo and provides alternatives. State authority isjustified when it prevents autocratic authority that is not responsible to thepopulace and subsequently overwhelms the populace’s autonomy and capac-ity to freely and fully associate with one another. Some defensive forms ofmilitary action can also be justified, although Chomsky maintains there is noevidence whatsoever to warrant the United States’s military interventions insuch places as Iraq, the Kosovo conflict, Vietnam, or Afghanistan. Generallyspeaking, state power or authority – particularly state military action – must bechecked by requiring it to satisfy a heavy burden of justification. Institutionsand practices cannot be defended solely by dogma; they can be determined tobe illegitimate based on their inability to meet the needs of individuals. He alsorequires that every powerful institution be regulated internally by meaningfuldemocracy, which consists of popular association, action, and control of allaspects of the institution, including management. These few, basic proceduraland substantive requirements demand that we continually evaluate a governmentand other forms of institutional authority by asking such empirical questions as:how decently does the institution treat each individual member (including those“aliens” unofficially affiliated with but nevertheless profoundly affected by thatsociety)? How are wealth and power distributed? How much choice do all peo-ple have in pursuing their chosen goals? How monotonous is their work? Whichof those individuals subject to the institution have meaningful, adequate rightsthat protect them from violence and from private and/or state tyranny?

Chomsky (1996b) distinguishes between his long-term anarchistic “vision”and shorter term “goals.” The latter range from exposing continuing oppression,eliminating the use of replacement workers (Chomsky & Barsamian 1998: 62),extending health benefits to all people (Chomsky & Barsamian 1994: 271–308), to enforcing basic human rights everywhere on the planet (Chomsky1999f). He rejects the view that people must suffer now in order to becomereceptive to revolution later, a ruthless Leninist argument that raises the specterof a replacement ruling class’s basing its legitimacy upon an alleged desire tohelp the people after they have accepted their place as subordinates. On theimmediate level, Chomsky advocates political reforms that are as incremental


248 James Wilson

as those sought by the remnants of the left wing of the US Democratic Party –slowly improving society through non-violent (Chomsky & Barsamian 1998:139), democratic action.

The corporation

One of Chomsky’s intermediate goals is to eliminate or at least reduce the pri-vate power of the wealthy few, a perpetual threat to stable and healthy societies.(He is particularly fond of quoting Adam Smith’s “vile maxim” of “the mastersof mankind”: “All for ourselves, and nothing for other people” (1996a,b: 77)).He finds the current system of state corporate capitalism “illegitimate” (1999c:146) and motivated by “inherent malice.” Corporations are instituted to maxi-mize profit and increase the power of the wealthy, not to serve the welfare ofall individuals. In this system, corporations are immortal “legal persons” thathave almost all the rights guaranteed to humans – short-lived moral persons.These rights are exercised to maximize profit. It is clear that the average citizenhas far less power than the corporate executive who can demand a meetingwith politicians and then exercise influence by threatening to withdraw futurecampaign donations. To Chomsky, the status of corporations as persons is anillegitimate legal fiction: “it’s hardly clear why such institutions should haveany rights at all” (1999c: 148).

Chomsky believes contemporary capitalism rests upon a dangerous ideology,a secular religion supported by a mythology for which there is little or noevidence. Consider, for example, corporate America’s commitment to “freeenterprise.” Over two hundred years ago, Adam Smith warned that the corporateform of economic organization allows particular interests to combine forcesto regulate prices. Chomsky concluded: “The pretence that corporations arenecessary for the better government of trade, is without any foundation” (1999c:148). Under this system, “efficient” means obtaining enough power withina particular market through financial profits to become immune from directcompetition and to demand substantial governmental subsidies. To a significantdegree, the economic concept of “economies of scale” now denotes “immunitiesfrom market forces” and “intimidation and control of employees.” The 1998merger of Mobil and Exxon demonstrates there is no limit to corporate effortsto concentrate wealth and power. Such concentration of power helps coordinatethe global market; it is hard for the titans of industry to convene if there aretoo many titans. This process places the fate of the planet in fewer and fewerhands, hands completely unaccountable to the average person except througha person’s very limited capacity to “vote with dollars.”

A little over a century ago, the American diplomat John Hay described theUnited States as “of the corporations, by the corporations, and for the cor-porations” (quoted in LaFeber 1967: 17). John Dewey, in turn, lamented the



shadow that big business cast over American culture (Chomsky 1999c: 146).Except for the Depression and its aftermath, the state has increasingly becomecapitalism’s best tool, particularly in the US. Chomsky has argued for yearsthat the Pentagon has little to do with national security, but everything to dowith profit (1993d: 93). By pouring money into “defense,” American taxpayershave financed advanced research into emerging technologies that are presentedas having some relationship to military needs. Once those products have beenrefined to the level that they can be profitable, whether as arms sales (in whichthe US leads the world) or commercial transactions like the development of theInternet, they are shifted over to well-connected parts of the private sector. Theprofits are not returned to the taxpayer whose funds made them possible, but arefunneled to the corporate sector. There are many examples of taxpayer aid thatenriches the corporation. The National Institute of Health, like the Pentagon,directs subsidies for research to pharmaceutical companies that profit from thispublic largesse. When the US gives foreign aid, it places conditions on its dis-bursement, requiring the recipient government to purchase American goods.Corporations receive “corporate welfare” in the form of tax breaks, exceptionsto regulatory principles, and lax enforcement of law.

Governmental discretion in its use of domestic force also benefits the wealthy.For instance, the Internal Revenue Service spends a great deal of its budgetchasing after the working class’s underreporting to gain an “earned income taxcredit” while steadily reducing its oversight of wealthy taxpayers’ returns. Alarge part of the domestic expenditure of the “war on drugs” budget consists notof efforts to improve conditions among the impoverished but increased policeand prison expenditures. The targets of these expenditures are drug users andpetty thieves – particularly those who are black – even though a large proportionof drug users are white, middle- and upper-class individuals. In contrast, there islittle scrutiny of corporate theft, such as consumer fraud, corporate accountingshenanigans, and antitrust violations, and little is done about corporate vio-lence in various forms – unsafe working conditions, dangerous products, andenvironmental contamination. When Congress makes a few minor reforms inresponse to the Enron/Worldcom scandals, the inadequately funded Securitiesand Exchange Commission subsequently diluted those efforts by watered-downregulations and lax enforcement. In recent years in the US, companies have beengiven increased access to low-wage prison labor, providing a pool of cheap tax-subsidized labor that requires no benefits. Industry profits from increased prisonpopulations in another way too, of course: increased populations necessitate theprivate construction and even maintenance of high-cost prisons.

The wealthy control the state in various ways, both directly and through theircorporate institutions. Lobbying and campaign financing are obvious examples.Both are difficult to control. The increasing clamor in the US for campaignreform, would, even if instituted, not eliminate the power of the rich. When


250 James Wilson

George Bush I was President, his son obtained many business “opportunities.”After the first President Bush retired, he quickly obtained profitable employ-ment in the private sector. It is important to remember that the wealthy havehad inordinate influence over American politics since the country’s inception.Equally difficult to regulate is the ruling class’s primary instrument of control,the well-entrenched corporate-run media. Whether it be drumming up hatredfor international villains (often servants of the American Empire who have for-gotten their places, such as Saddam Hussein), inundating the populace withpropaganda to reinforce the idea that the purpose of the corporate enterprise isto serve, and even redeem, the individual through consumption, or refusing topresent different viewpoints, private media have become the primary “doctrinalinstitution” (Chomsky 1996d: 162) for control of the populace. Note that forthis purpose it is not necessary to persuade the electorate to vote for a particularcandidate or even to vote at all. It is enough if people become convinced thatthere is little that average individuals can do. Judging by voter opinion pollsand diminishing voter turnout in the US, at least, corporate America has beenvery successful.

Corporate control of the state also yields international benefits. State-aidedcorporate control of international markets is often masked under the shibbolethof “free trade,” but close examination of any free-trade agreement reveals thatevery country attempts to maintain protective legislation for its more vulnerable,yet internally powerful, private interests. In actuality, “free trade” is power trade,in which the powerful seek state protection and powerful nation-states attemptto keep more “undeveloped” countries in their subordinate position. Chomskyprovides numerous examples of neoliberalism’s hypocrisy – Western nationsmaintain their own protective barriers while they demand that poorer nationsopen their borders to free trade, which allows the invasion of corporate power atthe expense of the average citizen and the environment (Chomsky 1993d: 33–64). For example, Mexican farmers are being wiped out by a combination of theNorth American Free Trade Agreement (NAFTA) and government subsidies toUnited States agribusiness. Industry then benefits from cheap labor, minimalunionization, and virtually no regulatory control.

The local gentry also benefits: some of the investment capital goes to them inthe form of government “aid,” generous loans, and bribes. Local governmentsfunnel cash and industry inducements towards speculation, reinvestment in theFirst World, and military and other expenditures to maintain control. Whenthe local markets crash, as they invariably have under this system, it is notthe wealthy elite and local government leaders who received the benefits whomust then pay, but the average person who is taxed to satisfy huge internationaldebts that “justify” even more severe economic “cures.” The resulting ThirdWorld “debt” is a political construct: the wealthy countries’ banks, who pro-vide the investment capital, demand that the poor country’s general population



assume legal responsibility for a rapacious local leadership’s poor investmentdecisions.

Those in the workforce suffer the greatest damage from the increasing powerof corporations. Often threatened with job loss should they protest or union-ize, workers perform grueling work for which they are paid less and less interms of absolute dollars and benefits. Over a period of decades, the steadyreduction in their disposable incomes, compounded by ever-increasing desiresfor goods (drilled into them by corporate advertising/propaganda departments),has precluded them from having sufficient time and energy to adequately raisefamilies, participate in their communities, and become involved in the polit-ical culture through unions and community organizations. One of the elite’sgreat advantages is the amount of time they have to organize and distributeinformation, whether through endless meetings, conferences, or golf outings.Meanwhile, producers are kept out of the decision-making process, reduced tothe narrow roles of worker and consumer. Their political participation is limitedto periodic opportunities to select a few members of the ruling class throughelections, members already culled by a process of campaign fund-raising thatallows the rich extraordinary access and control (Wilson 1996: 439). This elec-toral formality is a far cry from Tocqueville’s “Democracy in America.” Neitherfully human under Aristotle’s conception of the human as a political animal,nor happy under Humboldt’s definition of satisfying work, exhausted workerscan drift easily into political apathy reinforced by the bland compensations ofmass culture. Immersed in television and other forms of market-oriented “massculture” and subjected to a mediocre education (increasingly subject to corpo-rate intrusion), many citizens become disinterested in such “boring things” aslaw, politics, economics, and cultural management. Soon they are either notvoting or voting against their own interests, thus confirming the ruling class’sbelief that the average person’s incapacities warrant strong leadership.

Many people find it surprising that Chomsky places little importance on thepersonalities of individual leaders. He argues that there are more similaritiesthan differences among those who run powerful public and private institutions,because the leaders feel both internal pressures and external market forces thatreduce their discretion to accumulating more power for their institution pursuantto its enduring agenda. Such ruthlessness tends to require and attract ruthlessindividuals. Chomsky thinks that instead of creating a list of villains and heroes,it is better to study the system’s underlying purposes. Its dehumanizing con-straints are particularly noticeable in the corporate arena, affecting everyone –managers, investors, and workers. If a CEO actually wants to dramaticallyimprove the conditions of the company’s workforce, a goal that invariablycosts money, the shareholders would probably see that he or she is fired. At theleast, a more ruthless rival would probably purchase the relatively progressivecorporation to eliminate its alleged “inefficiencies.” The market is interested


252 James Wilson

only in profit: Wall Street downgrades any company that does not have – or isseen to have excellent prospects of having – a competitive return on equity. Thusallegedly impersonal, apolitical, and rational “market forces” preclude majorimprovements in the workplace. Mutual funds must show a profit, so they buystock only in those companies that return handsome profits. And the individual,cut out of the insider trading where much of the real wealth is distributed, mustcontend with a system that steadily reduces public services through “privatiza-tion” (as, for example, in education and government pensions2). The individualis thereby also driven to seek out the investments that are most likely to maxi-mize return.

Chomsky has characterized corporations as “private tyrannies” or “totalitar-ian” institutions (Chomsky & Barsamian 1996: 153). These shocking termsaccurately describe the constitutional structure of private corporations: powermoves from the top to the bottom. Those who work at the lower levels of thecorporation have no representation at the highest levels. Legally, they have nomore access to these institutions’ decision-making than the local community,the state, or the populace in general. Only the shareholders, generally con-sisting of a mixture of small individual investors and large corporate entitieswhose structures and ideologies are similar to those of the corporations theyown, can demand some accountability – although their power in this regardis determined solely by wealth because it is based on the number of sharesthey own. In the absence of protests by shareholders or the Board of Directors,the CEO effectively holds legislative, executive, and adjudicatory powers, aconcentration of power that James Madison once called “the very definition oftyranny” (Madison: Federalist no. 47). The typical CEO determines companypolicy, executes that policy, and decides which underlings have not adequatelycomplied with that policy. On a daily level, the corporate culture can approachtotalitarianism. Admittedly, some corporations are more tolerant and casualthan others, but that tolerance can be retracted at any moment. Furthermore,an easy-going atmosphere can further advance corporate aims: many high-techcompanies don’t care how people dress as long as they work over sixty hoursper week.

Many individual leaders of particular corporations will make such seriousmistakes that they will destroy their previously potent companies, but most willuse their power well enough to keep the overall system working. Collectively,private corporations are very good at maximizing profit. Those profits constitutean ever-growing source of capital, which enables the corporations to purchasemore goods and services from both the private and public sectors. One of themajor problems faced by the worker-owned factories that Chomsky advocatesis their likely inability to compete effectively against single-minded corporateentities. Humane institutions of production face the same dilemma as anar-chistic political societies contending with more aggressive rivals. Progressive



systems may provide better long-term solutions for most of the populace, butthey are vulnerable to predators. An anarchosyndicalist factory that cares aboutits workers will not be as financially strong in a “market”-driven society asits authoritarian rivals focused on the short term. The powerful drive out thedecent, forcing a person either to try to get rid of all the power at once (avery unlikely and risky enterprise with unforeseeable effects) or to accept thestatus quo. Capitalist market norms also pressure successful worker-run orga-nizations into either selling out or exploiting outside workers to maintain profitmargins.

Dismantling corporate power and reducing concentrated wealth are just stepstoward Chomsky’s greater vision of a humane mix of anarchy, socialism, lib-ertarianism, and democracy. One need not accept his longer-term anarchisticvision of eliminating the state to learn from him. Some believe that some sortof elite hierarchy is both inevitable (based upon differences in individual skills,fortunes, and motivations) and even desirable (to produce everything from stim-ulating art to good health care). As a reluctant defender of the need for somedegree of elite power, I think that the more realistic solution is to constrain “thefew” so that they cannot abuse their supplemental amount of power by plun-dering the majority. For those who worry that worker-owned factories might befar less productive institutions than those under management who aim towardsprofit, a less dramatic alternative is to limit capitalism through social democraticprinciples, hoping to obtain capitalism’s flexible, wealth-generating capacitieswithout suffering from its more pernicious traits.

A more humane political culture would offer state subsidies to allow worker-run institutions the time necessary to demonstrate that they can produce desiredgoods without degrading the environment or becoming as oppressive as thetraditional private corporation (even if they never become as “efficient” or“profitable”). Such experimentation should not be dismissed as an unprece-dented interference with the “free market” or as “socialism” – there is notmuch of a free market, particularly at the corporate level. Corporations aremajor recipients of public largesse. In particular, the accounting methods behind“profitability” need to be revisited. “Profit” currently is limited to the amountof wealth a corporation has managed to accumulate within corporate-internalaccounts. Other interests’ social expenses are excluded. For instance, a cor-poration might increase its profits by ten million dollars if it moves a factoryfrom one city to another while simultaneously inflicting fifty million dollars ofdamage on a local community. But if the workers own the factory, it would notbe moved and there would be less overall social cost. Moreover, corporationsusually relocate because the new community is offering tax incentives, greaterfreedom from environmental constraints, and the like – resulting in avoidableshort- and long-term damage to the population that “benefits” from the move.Very often, the total received benefit to the community in the form of wages


254 James Wilson

and tax revenues does not equal, let alone exceed, the amount of corporatewelfare provided to entice the corporation to relocate.3 Including the costs ofsuch “externalities” in private corporations accounts would reveal that they areby no means as efficient as they claim to be.

Capital mobility is the major corporate weapon used to reduce wages andextort state subsidies. Reducing capital mobility would weaken private power,increase public power at the local and national levels, and support alternativesbased upon more cooperative, community-benefiting norms. As Chomsky andothers have pointed out, this is not a radical proposal. Until recently, it was anaccepted economic principle. The economist John Maynard Keynes recognizedthat free-flowing capital undermines civilized welfare states. Keynes success-fully advocated imposing major constraints on the flow of international capitalby creating the Bretton-Woods system. These constraints had a major effecton the global market until President Nixon opted for “open markets.” It is notcoincidental that the average American worker’s income improved dramaticallyfrom the end of World War II until the early 1970s. Since then, there has beenvirtually no real increase.

The state

The Marxist vision has been partially validated: a ruling elite has relied upona materialistic, pseudo-scientific ideology with worldwide ambitions to witherthe state. The only problem is that the state is being replaced by a paradisefor investors and managers, not workers. Acting both individually and collec-tively as members of an international capital market, corporations are creatinga rival sovereign within each nation-state and an even more imposing world-wide system above each nation-state. In 1983, Harry Gray, the CEO of UnitedTechnologies described corporate utopia as “a worldwide business environmentthat’s unfettered by government interference,” such as “packaging and label-ing requirements” and “inspection procedures” to protect consumers (cited inChomsky 1996d: 163). According to him, government should be reduced to tworoles: providing corporate subsidies and protecting the wealthy from domes-tic and foreign threats. As long as the US government fulfills its role in thepartnership, it can count on the privately owned media to support it throughbiased news reporting that carefully documents every transgression by officialopponents but scarcely discusses the atrocities of Israeli politicians (Chomsky1983: 31) or South American dictators (Chomsky & Herman 1988: 37–86) whosupport the United States’ leadership.

Within every country, but especially the United States, big business is further-ing its agenda of redirecting public power so that its only obligation is to supportbig business (Chomsky & Herman 1988: 120–1). To provide further profits forexcess capital, necessary social services – such as schools, basic medical care,



prisons, and sanitation – are turned into opportunities for corporations throughthe allegedly efficient process of privatization. Even though the vast majorityof US citizens do not know the actual figures on how much less a single-payernational health system – like that in Canada – costs to operate, they have longwanted single-payer health care because they believe every American shouldreceive basic health care (Chomsky & Barsamian 1994: 226; Navarro 1994).But government listens to corporations much more than to people. The Clintons’“progressive” health care proposal included massive subsidies to large insur-ance companies to continue the inefficient, but lucrative, practice of boggingdoctors and patients down in paperwork. The Democrats appear to be somewhathumane only because their rivals, the Republicans, make little pretense of caringfor the average person. The Republicans obtain their twenty-plus percentageof the electorate (which is all it takes to win with so few people interested invoting) by pandering to religious and racist extremists. To avoid actual controlby these groups, which often have agendas opposed to increasing the wealthof corporations, country club Republicans donate millions to their favorites,or even to Democrats, who now resemble the conservative wing of the 1950sRepublican party. The resulting two-party system offers no room for genuinereform.

For decades, the rich and powerful have argued that the government, whichthey largely control, is inefficient, socialist, or worse. The media they controlstrips politics of its significance by reducing it to personalities and scandal.As Chomsky notes, all this might seem surprising until one recognizes that itis in the interests of the rich to denigrate public power. First, the governmentremains the most powerful potential threat to the corporations, just as the fed-eral government was once the greatest threat to Southern slavery. Second, thewealthy do not completely control the government, especially in formal politicaldemocracies like the US (Chomsky 1987b: 114) where, although insufficient,the electoral process permits people some access to power if they are willing toorganize and participate. In addition, formal state democracies have relativelylittle coercive power over their citizens; persuasion must be the primary toolof social control (Chomsky 1997b: 114).4 Private power must use its doctrinalinstitutions to convince the populace that there is no hope in politics and thatself-validation is available only through consumption. Private power’s flagrantabuse of its control of the government works to its advantage; disgusted bythe pervasive corruption, many people retire to their private life of family andfriends. The field is left open to those who are ambitious and greedy, willingto serve in the government as go-betweens between private power and a surlypublic.

There are, of course, limits to corporate consensus. Chomsky observes thatcorporate America often disagrees over particular issues: one segment sees theutility of some governmental regulation, if only to preserve their own interests


256 James Wilson

(Chomsky & Barsamian 1996: 149–51). And there has long been a portion of theruling class willing to share some of the wealth to guarantee political stability.The problem with relying on this kind of noblesse oblige, however, is that it canalways disappear. Aristotle (Politics 2010) warned of the elite’s insatiable greedand lust for power: “The fact is, that the greatest crimes are caused by excess,not by necessity.” For example, many members of the American plutocracywants to reduce Internal Revenue Service audits of taxpayers in high incomebrackets, turning the United States into a Russian-type system where only wageearners pay their legal share of taxes.5

The wealthy have not been satisfied with using corporate proxies, marketmuscle, and campaign donations to seize effective control of the United Statesgovernment. They have joined with their colleagues throughout the world toform a “de facto world government”6 to ensure the perpetual triumph of capi-talism (Chomsky 1993d: 51). The list of institutions promoting this goal growssteadily: the United Nations (Chomsky & Barsamian 1998: 114–17), the WorldBank, the International Monetary Fund (IMF), the North American Free TradeAgreement (NAFTA), the General Agreement on Tariffs and Trade (GATT), andthe World Trade Organization (WTO). Chomsky helped spread the news aboutthe proposed Multilateral Agreement of Investment, which would have madecapital flow immune from governmental regulation for many years (Chomsky1999c: 135). If it had been instituted, no one could have restrained capitalfrom entering countries with egregious human rights violations (1999c: 149).Investors and corporations could have sued countries for creating environmen-tal regulations that restrain the investors’ “enjoyment” of their “foreseeable”profits (1999c: 141). The WTO would have had legislative, executive, and adju-dicatory functions, staffed by private parties, with the power to trump the lawsof sovereign states. Nor is there any reason to think that the rich would havestopped with these “reforms.”

Faced with such overwhelming power, both private and public, it is under-standable that so many citizens prefer apathy to despair. After all, there is moreto life than politics. But as long as there is hope and the possibility of alternativeinstitutions, dissent offers great satisfactions. Chomsky argues that labor unionsand labor-based political parties must be part of this process. Until unions orga-nize the lowly paid, develop formidable political action groups, and providemoney and political organization to offset the disproportionate advantages ofthe rich, it is hard to envision significant change. But the goal is not just torevive union power, which could easily be coopted again. To achieve the sort ofsociety that will allow the full development of individual potential, the populacemust demand governmental support of alternative modes of production that arenot measured solely by profit.

Chomsky has provided a concrete example of an alternative that is a majorstep in the right direction: the Spanish worker-owned enterprise called the



“Mondragon Cooperative Corporation.” In 1998, it had over five billion dollarsin sales and seven billion dollars in assets within its subsidiary banking system(Chomsky & Barsamian 1994: 290–1). The management is accountable to acooperative congress and a standing committee. The cooperative congress con-sists of representatives elected from among workers in all the cooperatives in thesystem, with larger cooperatives receiving more representatives. The congresselects seventeen of its members to the standing committee, which appoints theCEO, approves the CEO’s major appointments, and is the equivalent of a boardof directors. Daily management is similar to that in conventional firms, butthe workers retain significant influence and control through their representa-tive system. The bank provides significant amounts of venture capital as wellas traditional consumer and business loans (Freundlich 2000). Originally pro-tected from foreign competition by Spanish laws, Mondragon now faces seriouschallenges as it competes against multinational corporations more willing tosacrifice the well-being of their workers.

Chomsky does not oppose state intervention in the economy. He notes thatthose countries with “state-guided development,” such as Japan and Korea,which have not completely subordinated their citizen’s interests to the wealthyor to the global market, have done much better than those nations that havetotally capitulated to corporate Washington’s demand for “free trade” (Chomsky1996d: 155). During a radio interview, a listener asked: “One of the things I’vealways thought, and I know this is probably not democratic, is why there isnot a limitation on the amount of profit anyone can make, any corporation, anybusiness?” Chomsky replied: “I think that [such a limit] is highly democratic,in fact. There’s nothing in the principle of democracy that says that powerand wealth should be highly concentrated so that democracy becomes a sham”(Chomsky & Barsamian 1994: 195). This argument, like so many others, canbe traced back to Aristotle (Politics 2008), who feared that large disparities inincome would cause political chaos by creating large numbers of poor individu-als, for “poverty is the parent of revolution and crime.” Aristotle recommendedlimits both to the amount of wealth a person would have and to the lust for moregoods: “[T]he legislator ought not only to aim at the equalization of properties,but at moderation in their amounts” (Politics 2010).

If the world shifted from private, profit-hungry corporations to syndicatesowned and managed by workers, would the state be eliminated? Chomskypoints out that even if we assume that a worker-owned factory would treat mostof its employees better than a private corporation would, we cannot be confidentthat it would act fairly toward all its employees. A majority of workers mightexclude people of a certain race or ethnic origin or assign people of a certaingender to low-paying jobs. They might be indifferent to such externalities asenvironmental damage. Also, to produce their goods, they have to trade withother syndicates, which requires some sort of externally enforceable system


258 James Wilson

of property and contracts. If Hobbes’s reading of history is at all reliable, onecan assume ambitious leaders will seize control of some of these syndicatesby claiming, perhaps accurately, that everyone in the organization will benefitif it becomes more aggressive. Some external agency is required to stop suchpredatory behavior, just as antitrust provisions should be currently used tostop the incredible consolidation of capital and power within a few privatecorporations.

Human institutions and the environment

Chomsky tends to ground his political analysis in human needs and interests.Consider one of his criticisms of GATT and NAFTA in 1994, a charge that isequally applicable to the newer WTO: “[P] roperty and investor rights are pro-tected in exquisite detail . . . while workers’ rights are ignored, along with therights of future generations (environmental issues)” (1996d: 164–5). Althoughthere is little doubt that the best way to persuade humans to change their behav-ior is to appeal to their own interests, including the interests of their progeny,it may be necessary to acknowledge that the rest of the ecological system hasworth independent of human rights. Although Chomsky recommends insti-tutions based on the humanistic philosophies of the ancient Greeks and theEnlightenment, we also must create alternative institutions and norms that tran-scend the humanist tradition, which is focused on our species alone. PerhapsChomsky is heading in this direction: his commitment to “sustainable growth”implies an economic system that can sustain both the species and the planet. In1999, he expressed sympathy for the norms underlying vegetarianism (althoughcontinuing to eat meat). Perhaps it is bourgeois sentimentalism to try to see theworld through the eyes of other animals, but it is essential to move beyond theexisting pillage of both humanity and nature to a world where people treat eachother decently without denuding the planet in the process. Admittedly, most ofus will identify far more with the spotted owl, the harp seal, and the salmonthan with the cockroach and the AIDS microbe, but there is great fulfillment inidentifying with life processes beyond one’s species, just as there are enduringrewards in caring for humans other than oneself.

So what does Chomsky think an individual should do? First, turn off thetelevision for as long as possible, at least until you have read enough to determinehow the elite are managing the planet. Distrust political assertions unsupportedby facts and clear argument. Look at the issues from many perspectives. Hesuspects you will emerge from this process profoundly disturbed by the extentof systemic injustice. The next step is to liberate yourself (assuming you have anincome that meets your basic needs) from the circle of consumption to permitmore time for family and friends. After that, talk to others and commit yourselfto some form of public action, whether union organizing, demonstrating, writing



letters on behalf of political victims, or teaching someone how to read. Relearnthe pleasures of working with others to make this planet a better place for theaverage person. Become part of an alternative world vision not based purelyupon accumulation and domination. As Chomsky put it: “Pick your cause and govolunteer for a group that is working on it” (Chomsky & Barsamian 1998: 152).Above all, never give up hope – for yourself, your country, your remarkablespecies, your planet.


13 Noam Chomsky: the struggle continues

Irene Gendzier

Introduction

“. . . no great power – not even one so selfless and beneficent as the UnitedStates – has the authority or the competence to determine by force the socialand political structure of Vietnam or any other country, no right to serve asinternational judge and executioner” (Chomsky 1971b: 86). Noam Chomsky’sposition was articulated with a characteristic blend of irony and blunt eloquencein his 1971 lecture in honor of Bertrand Russell. It was accompanied by hisaffirmation that “the radical reconstruction of society must search for ways toliberate the creative impulse, not to establish new forms of authority.”

To these guiding principles of Chomsky’s political thought, a third may beadded, a commitment “to penetrate the clouds of deceit and distortion and learnthe truth about the world, then to organize and act to change it” (1996a,b: 131).It is both a statement of purpose and a strategy for action. It defines Chomsky’spolitical work – his ongoing critique of US foreign policy, and his relentlessexpose of “the clouds of deceit and distortion” designed to mask it. The resultis a record of historical retrieval that has fundamentally altered what we knowof US foreign policy. The Chomsky files testify to the power embedded in suchretrieval. To engage in retrieval is to question who we are and what we know.It is to explain why we have accepted a distortion of history that is akin to itstheft. And it is to face the responsibility of change.

As Chomsky’s work has demonstrated, the key to understanding this distor-tion lies in uncovering what is concealed by privileged elites who protect theirinterests by deflecting public scrutiny and masking their policies by creatingthe “necessary illusions” that are needed in a democracy. Edward Said, in hispreface to The Fateful Triangle, described the contrast between true accountsof policy and their counterfeit as a “protracted war between fact and a series ofmyths” (Chomsky 1999b: vii). It is a war for the legitimation of policy that isfought in a democracy without resort to arms, coercion or formal censorship.The elites’ weapons of choice are those of ideological control operating throughthe collaboration of the media and compliant intellectuals who uncritically –though not unwittingly – serve the interests of the state.

260


Noam Chomsky: the struggle continues 261

The theme dominates Chomsky’s work. It explains his emphasis on exposingthe legitimation of policy and its roots in what Edward Bernays approvinglydescribed as “the engineering of consent” (Chomsky 1982: 66). The manipu-lation of opinion is an integral part of the public policy of democratic statesallied to corporate power. Those responsible, the courtiers of state and corporatepower, represent a “caste of propagandists who labor to disguise the obvious,to conceal the actual workings of power, and to spin a web of mythical goalsand purposes, utterly benign, that allegedly guide national policy” (1982: 86).

The consequences of their actions are far-reaching. They shape the param-eters of public debate, marginalize opposition and trivialize dissent in accordwith the interests of those with power and influence in government and theprivate sector. The implications for the domestic legitimation of US foreignpolicy are difficult to exaggerate. “It would have been impossible to wage abrutal war against South Vietnam and the rest of Indochina,” Chomsky andHerman observed in 1988, “if the media had not rallied to the cause, portrayingmurderous aggression as a defense of freedom, and only opening the doors totactical disagreement when the costs to the interests they represent became toohigh” (Chomsky & Herman 1988: xv). Manufacturing Consent offered compa-rable evidence for the cover-up of US policy in Central America in the 1980s.Earlier, Chomsky had pursued the question in his unmasking of US policy in theMiddle East after 1967. Furthermore, the end of the Cold War, as he frequentlyobserved, in no way altered the practice, as the examples of the coverage of theGulf War and that concerning US policy in Kosovo indicated, and as Chomskyshowed in his expose of the self-congratulatory rhetoric of the advocates of anew “military humanism” (1999e: 8).

The betrayal of intellectuals who have a responsibility to speak the truthremains a persistent current in Chomsky’s work. It reflects his view of intel-lectuals as a separate and privileged category, including that of “ideologicalmanagers” (1987b: 53). It is a role that is all the more revealing in states – suchas the United States – where intellectuals are free to act “without undue fear ofstate terror, to bring about crucial changes in policy and even more fundamentalinstitutional changes” (1985: 1). As he wrote in 1980, “if we do not like what wefind when we look at the facts – and few will fail to be appalled if they take anhonest look – we can work to bring about changes in the practices and structureof institutions that cause terrible suffering and slaughter” (1982: 337). Hence,the culpability of those who echo deceptive formulas “used to justify the nextdefense of freedom” is all the more flagrant, in Chomsky’s view (1969c: 359).

It was in the midst of the Vietnam War that Chomsky argued that the time hadperhaps come to conceive of such an alternative, a society “in which freely con-stituted social bonds replace the fetters of autocratic institutions” (1987a: 153).The humanist vision of that society reflected the origins of Chomsky’s politicalthought in anarchism, Enlightenment philosophy, and classical liberalism, all


262 Irene Gendzier

of which helped shape his views of the state and of “predatory capitalism” –and resistance to it represented in struggles for equality and social justice.

Chomsky’s critique of US policy assumed a vision in which the “sight of‘ordinary people, fighting for their liberty and independence with courage andintegrity,’ ” might one day be realized in practice (1988a: 773). This image wastaken from Rousseau, but Chomsky’s view of an alternative form of societyalso borrowed from Bertrand Russell and John Dewey, among others. Such asociety would rest, he reflected, on the ideas of equality and democracy, not thoseof “accumulation and domination, but independence of mind and action, freeassociation on terms of equality, and cooperation to achieve common goals,”he wrote. The impact of such ideas would be revolutionary, he maintained,because people “would share Adam Smith’s contempt for the ‘mean’ and ‘sordidpursuits’ of ‘the masters of mankind’ and their ‘vile maxim’: ‘All for ourselves,and nothing for other people,’ the guiding principles we are taught to admireand revere, as traditional values are eroded under unremitting attack” (1996a,b:77). The Adam Smith of conventional accounts was not the one that Chomskycited in such interpretations. It was, on the contrary, Smith’s “rather nuancedadvocacy of markets,” to which Chomsky referred, an advocacy based “inpart on the belief that under conditions of ‘perfect liberty’ there would bea natural tendency towards equality, an obvious desideratum on elementarymoral grounds.”

The struggle against the world of “sordid pursuits” was heavily compro-mised, in Chomsky’s view, by the postwar alliance of corporate capitalismand the state. That alliance represented the conjuncture of economic, political,and ideological systems increasingly “taken over by vast institutions of pri-vate tyranny that are about as close to the totalitarian ideal as any that humanshave so far constructed” he argued (1996a,b: 71). Exposing and resisting suchforces led him to adopt short-term goals obviously at odds with anarchist ideals.But Chomsky maintained that strengthening “elements of state authority” wasessential to preserving the considerable achievements of classical liberalism,namely, the extension of democracy as well as civil and political liberties(1996a,b: 73). Those achievements were threatened by the resurgent neoliber-alism of the post-Cold War period, with its heady endorsement of privatizationand globalization.

What then is to be done? Criticized for his pervasive pessimism, Chomsky infact has not wavered in his insistence on the possibility of social and politicalchange through collective action, albeit action that requires clarity with respectto values as well as a supportive institutional structure. The voluntarism implicitin his outlook is reflected in the position that “socioeconomic orders are humancreations, and to that extent they are subject to change,” he reminded listenersin New Mexico in the winter of 2000.1 The same conviction was articulated ear-lier in his belief that the “drive towards intervention, militarization, increased



authoritarianism, submissiveness to the doctrinal system, and possibly even-tual nuclear destruction is the result of human decisions taken within humaninstitutions that do not derive from natural law and can be changed by peoplewho devote themselves to the search for justice and freedom” (1982: 215). Thatsearch assumed a clear understanding of the roots of American power that wasa prerequisite to making sense of the world – and to changing it. But change, inthe direction of “meaningful democracy,” had other prerequisites, namely, “anorganizational structure that permits isolated individuals to enter the domain ofdecision-making by pooling their limited resources, educating themselves andothers, and formulating ideas and programs that they can place on the politicalagenda and work to realize” (1987b: 123).

Understanding the world

From his earliest writing on the Vietnam War Chomsky sought to map Americanpower and to identify the official “reasons of state” and their implicationsfor US policy in Southeast Asia and, by extension, much of the Third World(1973b). The results persuaded him of the importance of identifying the postwarfoundations and recurring principles and patterns of US policy. He locatedthe evidence in theory and practice, in US internal documents and the evidenceof US policy. The conformity between the two was not in doubt. What theevidence of internal sources disclosed, however, was the gap between insidetalk and that designed for the public, a gap reinforced by academic apologiesand media echoes of government policy. The real story, for those seeking tounderstand US policy, as Chomsky’s work made clear, was in part to be foundin official sources, records of policy planning, such as the Pentagon Papers, andthe reports of think tanks dominated by US political and business elites, suchas the Council on Foreign Relations with their insights into postwar policy.These sources confirmed the connections between private and public sectors,between “representatives of major corporations, banks, investment firms, thefew law firms that cater to corporate interests, and the technocratic and policy-oriented intellectuals who do the bidding of those who own and manage thebasic institutions of the domestic society” (1982: 91).

The influence of such sectors marked the nature of postwar US policy with itscomplex of economic, political, and military corollaries. At one level it involvedwhat Chomsky termed the “first principle,” the commitment to “maintain aninternational order in which U.S.-based business can prosper, a world of ‘opensocieties,’ meaning societies that are open to profitable investment, to expansionof export markets and transfer of capital, and to exploitation of material andhuman resources on the part of U.S. corporations and their local affiliates”(1987b: 6). At another, it involved the political and military ramifications of suchpolicies in the form of the containment of radical options deemed hazardous


264 Irene Gendzier

to US interests. The results were reflected in the identification of key zonesof industrial and political reconstruction that functioned as pivots of Americaninfluence and in the accompanying restoration of traditional political forcesdesigned to assure the containment of radical change. They were illustrated insuch landmark policies of the postwar decade as the Marshall Plan, the TrumanDoctrine, National Security Council Memorandum 68 with its militarized visionof unrestricted rollback policies, and the less frequently cited (1947) merger ofUS oil multinationals in ARAMCO, with its obvious importance to global USpolicy, especially for the Mediterranean and Middle East.

On this and the more general question of postwar US policy and the ColdWar, Chomsky broke with dominant interpretations long before the collapseof the USSR and the communist regimes of Eastern Europe. He regarded thestandard explanatory use of the Cold War as entirely inadequate (see 1996d:ch.1, §4). He charged both major powers with exploiting Cold War doctrine tojustify their respective policies. For the former Soviet Union, Chomsky main-tained, “the Cold War has been primarily a war against its satellites” (1991:28), a war “illustrated by tanks in East Berlin, Budapest and Prague, and othercoercive measures in the regions liberated by the Red Army from the Nazis,then held in thrall to the Kremlin; and the invasion of Afghanistan, the onecase of Soviet military intervention well outside the historic invasion routefrom the West” (1991: 20). On the domestic level in the USSR, the Cold War“served to entrench the power of the military-bureaucratic elite whose rulederives from the Bolshevik coup of October 1917.” In the case of the US, onthe other hand, Chomsky argued that the Cold War represented a “war againstthe Third World” (1991: 28). It was a war that entailed a “history of worldwidesubversion, aggression and state terrorism,” accompanied at the domestic levelwith “the entrenchment of Eisenhower’s ‘military-industrial complex,’ ” whichsignified a “a welfare state for the rich with a national security ideology forpopulation control (to borrow some counterinsurgency jargon), following theprescriptions of NSC 68” (1991: 21). Grand Area foreign policy (the US usingmilitary and economic power to control markets in as large an area as possible)would be wedded to a military-industrial complex at home. Market ideologywould, in effect, drive US foreign and domestic policies.

In Chomsky’s analysis, the Cold War was subordinated to another, thatbetween North and South. In his scenario, the East–West conflict did not dis-appear, it was redefined to underscore the Western fear of the appeal to ThirdWorld nations of the Soviet Revolution of 1917 – fear that citizens of develop-ing nations might take their economic affairs into their own hands. The ColdWar was thus integrated into the Other War, that between an industrializedNorth and the global South, with its “resources, cheap labor, markets, oppor-tunities for investment” (1993d: 33). The major themes in the battle of Northagainst South are “imperialism, neocolonialism, the North–South conflict, core



versus periphery, G-7 (the leading state capitalist industrial societies) and theirsatellites versus the rest” (1993d: 3). Considered in broad historical perspec-tive, the Cold War represented “an interlude in the North–South conflict ofthe Columbian era, unique in scale but similar to other episodes in significantrespects” (1993d: 65). From this vantage point, the collapse of the USSR did notsignify the end of the Cold War redefined as a war against autonomous, demo-cratic development. The policies of the industrial powers, Chomsky pointed outin a symposium on the question, “are adapted to changing circumstances, ofwhich the end of the Cold war is only one aspect” (1992a: 85).

Chomsky’s reconceptualization of the Cold War effectively established theprimacy of the struggle for resources and markets as the principal postwarobjective of US policy, an objective whose implementation necessarily involveda broad array of policies with a global reach. In this context, the future ofdecolonization struggles aroused obvious concern, and not only in Washington.US policy-makers – as well as their British and French counterparts – viewedthe emergence of nationalist movements in the Third World as a development tobe managed and redirected to serve their respective interests. The results wereapparent in French, British, and US policies in Asia, Africa and the Middle East,and Latin America. They were apparent in the overt and covert interventionsdesigned to roll back efforts of nationalist, populist, and radical regimes toregain control over their own political as well as economic resources. Statesand societies that sought an independent course were severely punished, lesttheir example inspire others, as Chomsky and Herman demonstrated in ThePolitical Economy of Human Rights (Chomsky & Herman 1979a,b), and asChomsky illustrated in countless works on postwar US policy in Europe, Asia,Latin America, and the Middle East.

The evidence of US policy Chomsky compiled confirmed the recurring pat-terns and underlying principles of US policy at work in such dissimilar envi-ronments. It was a record completely at odds with official endorsement of self-determination and claims to promote modernization and democracy. Instead,as Chomsky’s files revealed, the record of US policy involved the subversion ofdemocracy, the undermining of independent development and the legitimationof force in the Third World, in the name of democracy, development and therule of law in the First.

Records of intervention

Vietnam

It was the US war in Indochina, the ferocity of the US destruction ofSouth Vietnam, the “secret war” against Laos and the infernal destruction ofCambodia, the raining of death in the absence of war, that propelled Chomsky


266 Irene Gendzier

into the role of indefatigable critic of US policy (1969b: 60–1). His writingson Vietnam have been described as “among the most valuable ever written pre-cisely because they show[ed] so much of the war’s reality at the time, far morethan most of the current outpouring of books reassessing the war’s meaningtoday” (1987a: xv).

That reality concerned the US invasion of South Vietnam in the name ofanti-communism and the infliction of mass suffering through campaigns offorced relocation and resettlement justified in the name of modernization. “TheAmerican war in Indochina,” Chomsky wrote in 1969, “has been based on twoprinciples: physical destruction in areas that are beyond the reach of Ameri-can troops, and the use of what are euphemistically called ‘population controlmeasures’ in areas that can be occupied by American forces or the forces thatthey train, supply, advise, and provide with air and artillery support” (1969b:52–3). What this meant in practice was “massacre and forced evacuation of thepeasantry, combined with rigorous control over those forced under Americanrule.” That, Chomsky argued, “is the essence of American strategy in Vietnam.”

Washington moved first to support and then to replace the French when theywithdrew from Indochina in 1954, the year in which Washington underminedthe Geneva Accords with their promise of elections and the prospects of reuni-fication, and moved to take control of the Ngo Dinh Diem regime which itprimed in pursuit of its rollback efforts. By that time the US was already “pro-viding about 80 percent of the costs of the war” in Indochina and was close todeploying nuclear weapons (1987a: 323). Years earlier the State Departmenthad recognized “that Ho Chi Minh was the sole significant leader of Vietnamesenationalism, but that if Vietnamese nationalism was successful, it could be athreat to the Grand Area, and therefore something had to be done about it”(1987a: 322). Vietnam’s principal threat, in short, was the feared “dominoeffect” of its policies. It was met in 1962 by the “bombing and defoliation inVietnam” by the US Air Force,

part of the effort to drive several million peasants into concentration camps where theycould be ‘protected’ from the guerrillas who, the government conceded, they werewillingly supporting, after tens of thousands had been slaughtered and the United Stateshad effectively blocked any political settlement, including the offer of the NLF (the‘Vietcong’ in terms of US propaganda) to neutralize South Vietnam, Cambodia, andLaos. (1987a: 52–3)

The bombing of South Vietnam assumed a new ferocity between 1965–9.Pentagon sources revealed the deployment of “nine times the tonnage of bomb-ing in the entire Pacific theater in World War II, including Hiroshima andNagasaki – ‘over 70 tons of bombs for every square mile of Vietnam, Northand South . . . about 500 pounds of bombs for every man, woman and child inVietnam’ ” (1969b: 291). The figures were incomplete insofar as their account



of military action or the toll of intimidation, repression, and torture by the US-backed South Vietnamese regime, as well as the US massacres of which MyLai was but a single example of a far more widespread phenomenon.

Chomsky’s response was unequivocal. “With no further information thanthis,” he wrote referring to the above, “a person who has not lost his senses mustrealize that the war is an overwhelming atrocity” (1969b: 291). The positionwas not widely shared in the US. In fact, it was widely denounced by criticsof the war, let alone its supporters. Mainstream critics of the Vietnam Warremained well within the acceptable bounds of criticism, formulating theirregrets in terms of the inevitable limits of American benevolence or the cost-benefit analyses of an increasingly expensive war. In such circles, as Chomskypointed out, those who refused to accept the “fundamental axiom, namely, thatthe United States has the right to extend its power and control without limit,insofar as is possible,” were regarded as “hysterical critics” whose irrationalviews were beyond debate (1969c: 333). Viewed as deviants, their real crime, asChomsky understood, was to defy apathy and to resist and oppose US policies.The “civic culture” defended by the “secular priesthood” could not toleratesuch behavior or the threat of its extension to liberation movements across theThird World, including those emerging in the Middle East and Latin America.They were tarred with the label of the “Vietnam Syndrome,” reserved for thosewho exhibit the aversion to “massacre, aggression, and torture,” that reflects a“solidarity with the victims,” as Chomsky explained (1987a: 337).

Chomsky’s opposition to the Vietnam War extended to his denunciation ofthe distortion and deceit perpetrated by government officials and their academicacolytes, supported by the media. It was in this context that he called on intellec-tuals “to speak the truth and to expose lies,” with the knowledge that they werein a position to disclose “the lies of governments, to analyze actions according totheir causes and motives and often hidden intentions” (1969c: 324). He calledon universities to provide refuge to students and to assist them in defendingthemselves from “massive government propaganda apparatus, from the naturalbias of the mass media, and . . . from the equally natural tendency of significantsegments of the American intellectual community to offer their allegiance” topower (1969c: 313–14). And he urged those willing to resist to publicize “thedaily evidences of barbarism, and the still more severe duty of challenging thepowers – state or private – that are responsible for violence and oppression,looking forward to the day when an international movement for freedom andsocial justice will end their rule” (1969b: 313).

Chomsky and Herman publicized the record of barbarism beyond Vietnamand their account of US actions in Laos and Cambodia showed how firmlythe US adhered to Grand Area policy. It exposed the systematic deceptionthat accompanied US policy; its veiled and distorted images rendered itsvictims “unworthy,” in accord with the policy of the victors (Chomsky &


268 Irene Gendzier

Herman 1979b). It is difficult to exaggerate the importance of their analysisand documentation of this phase of US policy, or the continuing controversyit has generated. The level and consequences of US bombings of Cambodia,the human, political, and physical devastation they caused, were ignored in theUS and Western media. Cambodia became a proper subject of media cover-age only in the period of Khmer Rouge “murderous rule” and “atrocities,” theauthors noted. The legitimacy proffered was self-serving. It stopped short ofconfronting the impact of US bombings and its obliteration of the prospects of“social and economic progress,” ignoring “the brutality of the eventual victors”(Chomsky & Herman 1979b: 219). That outcome, as Chomsky and Herman’swork demonstrated, was exploited in “a retrospective justification for earlierFrench and American crimes in Indochina, and [facilitated] the reconstructionof Western ideology after the Vietnam trauma . . .” (Chomsky & Herman 1988:296).

Chomsky returned to US policy in Vietnam and Indochina throughout hislater works as the US wars in Asia were reshaped “to preserve the image ofU.S. benevolence, always a crucial element in imperial ideology,” as Chomskyand Herman wrote in their 1979 account of the postwar condition of Indochinaand “the reconstruction of imperial ideology” (1979b: 10).

Latin America

Geographically distant, US intervention in Vietnam and Central America hadmuch in common. Chomsky said as much when he questioned whether USaction in Vietnam was

simply a single outburst of criminal insanity, of no general or long range significanceexcept to the miserable inhabitants of that tortured land. It is difficult, however, to putmuch credence in this possibility. In half a dozen Latin American countries there areguerrilla movements that are approaching the early stages of the second Vietnamese war,and the American reaction is, apparently, comparable. (1969c: 311)

The prognosis was accurate. The Vietnam interventionist policy was, ineffect, reproduced in different form and context in Central America, save thatthe public reaction to the Vietnam War in the US precluded support for directUS intervention in Central America, a matter of overwhelming importance, asChomsky insisted. Instead, there were to be proxy wars, doctrinally disguisedin terms of the Nixon Doctrine, the Reagan Doctrine, and other interveningdoctrines that assured that the US provided military and intelligence assistance,trained local forces in counterinsurgency, manipulated and funded politiciansin accord with its vision of how to stem radical change at any cost.

What marked Chomsky’s accounts of US policy in Latin America was notonly his documentation of the punitive history of US relations with Latin



America, as in the record of US intervention in Cuba, Panama, Mexico, Haiti,the Dominican Republic, Guatemala, Nicaragua and El Salvador, Costa Rica,Chile, Brazil, and Colombia, among other cases. It was, first, the extensive evi-dence of the extent to which US officials, fully cognizant of existing conditions,deliberately elected to support repressive regimes to avert those of a reformistor radical cast. And, second, it was the extent to which the media legitimizedsuch policies by their distortions or blanket omissions of their consequences,thus effectively purging them and their victims from the record.

US policies, Chomsky argued, illustrated Washington’s profound antagonismto struggles for democracy, to efforts to address issues of poverty and injustice.The lesson of US policy was that those committed to

the needs of the poor majority, or who seek to construct a political system that will not becontrolled by business-based groups and a military system not linked to and dominatedby the United States, are ‘Castros’ who must be driven to reliance on the Soviet Unionby unremitting attack, subjected to terrorist violence and other pressures, and crucially,prevented from perpetrating the crime of successful development in the interest of thepoor majority. (1987b: 89– 90)

US policy – under a succession of Democrat and Republicanadministrations – was driven, in practice, by the double commitment to assurefavorable conditions of trade and investment, guarantee access to raw materials,and promote export-oriented economies, while supporting compatible politicalregimes – irrespective of their repressive domestic policies, whose justificationin the US was invariably in anti-communist terms. The practice led to a pat-tern of political, military, and intelligence support, organized on a region-widebasis, with the “standardization of Latin American military organization, train-ing, doctrine and equipment along U.S. lines” (1987b: 20). In short, US policyconformed to the dominance of the “first principle,” the “Fifth Freedom,” andthe feared domino effect of independent and radical regimes.

The examples are too numerous to fully recite. The response of the US to theCuban revolution (1959) is perhaps the best known. The Kennedy administra-tion’s Bay of Pigs invasion was followed by policies that made Cuba “the primevictim of international terrorism for the next 20 years, probably surpassing therest of the world combined,” Chomsky adds, “if we exclude from the categoryof terrorism cases that might more properly be called outright aggression, suchas Israel’s bombing of Lebanon with US support from the early 1970s” (1987b:74). But there were other cases with regimes of a different cast, such as thatof Dominican President Juan Bosch, ousted from power by the military withUS support and then reinstated, only to be blocked once again with the supportof a military coup backed by the US dispatch of marines in 1965. Bosch was“a liberal democrat, committed to reformist capitalist democracy, meaningfuldemocracy with programs designed to serve domestic needs” (1987b: 65).


270 Irene Gendzier

Nearly a decade earlier there was the case of US intervention in Guatemala.In 1954, one year after the Anglo-American coup that ended the political careerof the Iranian Prime Minister who had dared to nationalize the Anglo-IranianOil Company, Eisenhower intervened in Guatemala to assure the collapse of theArbenz regime, punished for its successful land reform policies. In GuatemalaUS policies leading up to 1954 resulted in the collapse of the Arbenz regime andits reformist policies, decades of impoverishment for the indigenous popula-tion through support of terror and repression at the hands of US-trained militaryand intelligence agents, and an estimated 200,000 deaths. Chomsky and Her-man described the years that followed as those in which Guatemala “graduallybecame a terrorist state rarely matched in the scale of systematic murder of civil-ians, but its terrorist proclivities have increased markedly at strategic momentsof escalated U.S. intervention” (Chomsky & Herman 1988: 72). Human rightsgroups during the Reagan administration documented the same phenomenon,with “the indiscriminate killing of peasants (including vast numbers of womenand children), the forcible relocation of hundreds of thousands of farmers andvillagers into virtual concentration camps, and the enlistment of many hundredsof thousands in compulsory civil patrols” (Chomsky & Herman 1988: 73). Anestimated 100,000 civilians (1978–85) were murdered (Chomsky & Herman1988: 75), as the Guatemalan regime crushed indigenous human rights organi-zations, including those committed to discovering the “disappeared,” such asthe Mutual Support Group (GAM) (Chomsky & Herman 1988: 80).

Between the early 1960s and 1980 El Salvador witnessed the emergence of aproliferation of grassroots organizations as well as guerrilla groups committedto turning the tide against the progressive deterioration of living conditions in acountry ruled through the alliance of the landed elite and the military. The USresponse – accelerated by the Sandinista revolution in Nicaragua – endorsed a“preventive” military coup. It was quickly followed by a succession of intra-military struggles into which the US intervened, this time with military backingfor its candidate, Jose Napoleon Duarte, who presided over a slaughter of some40,000 victims when in power (1987b: 82). US military advisors followedwith training and support for the air war and the establishment of intelligencenetworks, backed from Honduras and Panama. Their targets were the advocatesof domestic reform. Signs of increased popular participation in widely supportedprotest movements designed to address the urgent needs of the impoverishedmajority, inspired alarm in Washington, where “church-based self-help groups,peasant associations, teachers unions,” among other opponents of the regime,were viewed as constituting an increasingly high risk to stability. The stabilitywas measured in terms of the interest of ruling elites; the risks were thoseof “meaningful democracy in which the population at large may be able toparticipate in shaping public policy” (1987b: 80).



As the military and security forces in El Salvador escalated the pattern of ret-ribution and violence, opposition intensified. Its victims included Jesuit priests,such as Father Rutilio Grande, the Archbishop of El Salvador, American church-women and countless others rounded up in the “state of siege” that reigned in1980. Archbishop Romero, who had personally requested of President Carterthat the US cease its support of right-wing terror, was murdered in the springof the same year. The trail of atrocities against peasants, journalists, universi-ties, libraries, and other signs of opposition accelerated. The response of the USmedia, as Chomsky and Herman reported, was a “virtual blackout” (Chomsky &Herman 1988: 47). Father Rutilio Grande’s murder was registered in the mediawith regrets, as it perpetuated the myth initiated by the US government, thatEl Salvador’s “moderate government” was “plagued by the terrorism of theextremists of the left and right,” which it was unable to control (Chomsky &Herman 1988: 59). Yet, as Chomsky and Herman argued, US officials as well asjournalists knew that those in control were the “U.S.-backed security forces,”as well as “the paramilitary network they created to terrorize the population.”Under the circumstances, they concluded, “these were crimes for which we bearconsiderable responsibility, since they were perpetrated by clients who dependon our support, so that exposure and pressure could have a significant effect insafeguarding human rights” (Chomsky & Herman 1988: 83).

In the case of Nicaragua, US clients were the counterrevolutionary exilesmobilized and supported by the US, the contras, whose purpose was to disablethe regime and abort the revolution. With the fall of the Somoza regime in1979, a major US ally, the Carter administration moved to assure its continuinginfluence over the US-trained Nicaraguan National Guard. Further, it focusedon strengthening the private sector with international financial support. Theobjective, as Chomsky pointed out, was to assure the failure of the regime. It was,he indicated in citing US former CIA analysts, to provoke repressive measuresthat would confirm the regime’s “allegedly inherent totalitarian nature and thusincrease domestic dissent within the country” (1987b: 83). Its achievementsin the areas of land reform, health, education, and popular participation, wereinternationally recognized, save in the US where Reagan sought to cut off aidand to pressure US allies in Latin America and Europe to follow suit. In short, theSandinista case illustrated US fears of the “demonstration effect” of successfuldevelopment in terms that might be meaningful to suffering people elsewhere,“endangering the Fifth Freedom,” as Chomsky explained (1987b: 84).

As a consequence, the Reagan administration intensified its efforts to assurethe political and economic destabilization of the regime. In 1983, Reaganauthorized the mining of Nicaraguan harbors, while in 1984, the CIA attackedoil tanks, communication and military centers to disrupt Nicaraguan trade.Chomsky reported on the international reaction in Necessary Illusions, and in


272 Irene Gendzier

other works, including his discussion of Rogue States, whose operations wereillustrated by the US response to the decision of the International Court. In June1986 the International Court of Justice

condemned the United States for its support for the contras and illegal economic warfareand ordered it to desist from the violations of international law and valid treaties andto pay reparations. The decision was reported, but dismissed as a minor annoyance. Itscontents were suppressed or falsified, the World Court – not the United States – wasportrayed as the criminal, and the rule of law was held inapplicable to the United States.

As Chomsky pointed out, the response of the major representatives of theUS media was to discredit the ruling, as “the US then vetoed a UN SecurityCouncil resolution calling on all states to observe international law, and votedin virtual isolation against similar General Assembly resolutions” (2000e: 4).

The Middle East

Chomsky moved to speak out on the Middle East in the midst of the VietnamWar. He faced the combined hostility of the US establishment, its liberal allies,and an anti-war movement that split on the question of Israel and Palestine. TheIsraeli–Arab war of 1967 had transformed the region. It underscored Israel’smilitary superiority over its Arab neighbors, led to Israel’s military occupationof the West Bank, the degradation of Palestinian life, the reorganization of thePalestinian resistance movement, and the radicalization of Middle East poli-tics. Washington’s response was a function of its primary interests in the region,uninterrupted access to the oil produced and transported by US multination-als, and the protection of conservative regimes on which US influence relied.Israel, enhanced by unconditional US political and military assistance, was anintegral part of the US regional system designed to contain Arab radicalism intandem with US regional allies. Israel’s overt as well as covert Arab policies –including its anti-Palestinian policies – conformed to US objectives as well as tothose of its conservative Arab allies, formally opposed to Israel but, in practice,beneficiaries of its anti-Palestinian and anti-radical policies.

Chomsky addressed these issues in an American context intolerant to criti-cism of Israel, one in which the dehumanization of Arabs and Palestinians, inparticular, promoted the distortion of Middle East politics and the legitimationof US and Israeli policies. As he observed with respect to the 1967 period,“topics that were widely discussed and debated in Europe or in Israel itselfwere effectively removed from the agenda here, and a picture was establishedof Israel, its enemies and victims, and the U.S. role in the region, that bore onlya limited resemblance to reality” (1999b: 29). Israel, it seemed, benefited from“a unique immunity from criticism in mainstream journalism and scholarship,consistent with its unique role as a beneficiary of other forms of American



support” (1999b: 31). As Chomsky wrote in 1983, “when the intellectual his-tory of this period is someday written, it will scarcely be believable” (1999b:1(n)).

The long-term impact of such privileged distortion is difficult to exaggerate.It accounts for Israel’s image as benevolent occupier, for the view of Palestini-ans as congenitally untrustworthy terrorists in waiting. It explains the prevailingignorance of Arab states and societies, their image as primitive backwaters ofresource-rich regions devoid of legitimate social and political movements, withthe exception of those few friendly “moderate” regimes whose main character-istic is their pro-American outlook. Struggles for democracy and social justicein the Arab world have no part in this vision, since to include them would exposeboth Israeli and US policies in actively seeking to undermine them.

Chomsky’s first full-length book on the subject, Peace in the Middle East(1974b) set out his views of the conflict and the conditions of its possibleresolution. It warned against Israeli expansion, military occupation, creationof settlements, exploitation of Palestinian labor, and the continued rejectionof Palestinian national demands (1999b: 20). And it exposed US support forIsrael’s occupation and its retreat from the internationally accepted interpreta-tion of UN resolution 242, a resolution that offered no recognition of Palestiniannational rights but that called for Israeli withdrawal from occupied territories.The shift in the US position, a matter of enormous consequence, as Chomskydemonstrated, was the product of a power struggle in Washington that resultedin the replacement of President Nixon’s Secretary of State – William Rogers –with Henry Kissinger. In terms of Middle East diplomacy, the abandonmentof the Rogers Plan (1969) was a prelude to military confrontation as opposedto diplomacy. Had this not been the case, Chomsky cites Israeli sources asconceding “that Israel could have had a peace settlement in terms of the pre-vailing international consensus, offering nothing to the Palestinians, by 1971”(1996d: 209). The foundations of Israeli rejectionism, as Chomsky consistentlyunderscored citing Israeli sources, was not security but territory and resourcesthat additionally explained the pattern of West Bank settlements. “The primarystrategic motivation for Israeli rejectionism, whether of the Likud or Labor vari-ety, is control over the territories and their resources,” a factor whose importancewas no less evident in the troubled course of negotiations between Palestiniansand Israelis after 1991 (1996d: 211).

Writing in the post-1967 period, Chomsky argued that Israelis andPalestinians were bent on “self-destructive and possibly suicidal policies, andthat, contrary to generally held assumptions, there were – and remain – alter-natives that ought to be considered and that might well contribute to a moresatisfactory outcome” (1974b: 10). Such alternatives included a “socialist bina-tionalism,” Chomsky maintained, as the “best long-range hope for a just peacein the region” (1974b: 210). It offered the most likely prospect of “reconciling


274 Irene Gendzier

the just and compelling demands of the two parties to the local conflict in theformer Palestine,” he suggested, recognizing that it was unlikely to be translatedinto practice immediately, he urged the importance of keeping “that hope aliveuntil such time as popular movements within Israeli and Palestinian society,supported by an international socialist movement that does not now exist, willundertake to make such a hope a reality” (1974b: 38).

Chomsky’s position on the core issues in the Israel–Palestinian conflictremained steadfast. He continued to uphold “the valid claims of those whoregard the former Palestine as their home” (1999b: 39), clearly reiterating

the principle that Israeli Jews and Palestinian Arabs are human beings with human rights,equal rights; more specifically, they have essentially equal rights within the territory ofthe former Palestine. Each group has a valid right to national self-determination in thisterritory. Furthermore, I will assume that the State of Israel within its pre-June 1967borders had, and retains, whatever one regards as the valid rights of any state within theexisting international system.

The statement appeared in 1983, in the aftermath of the Lebanon War, widelyanticipated by US, Arab, and Israeli critics of the Camp David Agreements(1979) between Israel and Egypt that effectively freed Israel to pursue its anti-Palestinian policies in Lebanon, where it simultaneously struck out againstthe Lebanese opposition to assure the restoration of a pro-Israeli Maroniteleadership. Israel’s Lebanon policy backfired with disastrous consequences forLebanese and Palestinians.

The immediate cause for Israel’s invasion of Lebanon, Chomsky observed,was not in doubt in Israel. It was the increasing international legitimacy ofthe PLO, a factor effectively “rated ‘X’ in the United States” (1996d: 214).The long-term cause of the invasion was no less in evidence, as Chomskydemonstrated by discussing its history and aftermath in his The Fateful Triangle,the United States, Israel and the Palestinians (1983, 1999b). The Israeli attackon Lebanon was designed to crush the Lebanese base of the PLO, therebyfatally undermining its influence in the West Bank. The US endorsed the planin a much-discussed interpretation of the “green light” that Washington offeredTel Aviv. In practice, President Reagan supported Prime Minister Begin’s policywith its demolition of the Lebanese opposition and the crushing of the PLO. Thedisastrous results of the invasion, with its destruction of Beirut, culminated inthe Israeli supervised massacre of Palestinians in the Sabra and Shattila refugeecamp, carried out by Israel’s Lebanese Phalangist allies.

Chomsky approached the Palestinian massacre and the question of Israeli aswell as US responsibility with an account of another massacre, that of the Jewsin Kishinev in 1903. The memory of the earlier massacre framed the questionof responsibility in the later one, and in so doing drew a wordless analogythat defied the dehumanization of its victims while exposing the crimes of its



victimizers (1996d: 332). The international reaction to Sabra and Shattila wasoutrage. The Kahan Commission Inquiry in Israel investigated the responsibil-ity of the Israel Defense Forces, leaving no doubts as to its implicit conclusions,as Israeli critics conceded (1996d: 408). US officials disclosed that the Israelimilitary had done “nothing to stop the carnage,” as Chomsky indicated (1996d:365). US intelligence was implicated as well, since US intelligence, accordingto Israeli press reports cited by Chomsky, had provided “hard intelligence infor-mation . . . confirming that Israeli military officers in Beirut were well awareof the brutal killings many hours before the Israeli Defense Forces actuallywent into the camps. . . .” For the then ex-deputy mayor of Jerusalem, MeronBenvenisti, “everything that has happened in Israel until now [Oct 4, 1982] hascarried the stamp of American approval, or at least, it was tolerated by yourgovernment” (1996d: 391). That approval, Chomsky argued, was a function of“ ‘ideological support’ for Israel in the United States, with its systematic fal-sification of the historical record and its practice of defaming the Palestiniansand ignoring their torment,” he argued (1996d: 393).

There is much that could have been done to present a fair and honest picture of what wasand had been happening, and to change the U.S. policies that have predictably led to therise of a Greater Israel that is a threat to its own citizens, to those subject to its militarypower, and to many others as well, and that lie behind the specific events of 1982. Tothe extent that we do not do what can be done, we have only ourselves to blame for theconsequences.

The result of the Lebanon war and continued Israeli repression in the WestBank was revolt, the Palestinian Intifadah of 1987. Its impact extended to thePalestinian political class that would later challenge the authoritarian leadershipof Yasser Arafat. In the interval, the Intifadah led to Israeli military reprisals,violence and savagery that escalated to what an Israeli Foreign Ministry officialclaimed was “becoming ‘a real war,’ ” one dutifully kept off American screensand more pointedly, “from the eyes of the American taxpayer who funds it, afurther contribution to state terror,” as Chomsky wrote (Chomsky 1989: 214;Macedo 2000: 75).

The end of the Cold War, while appearing to herald a new age in the ongoingconflict, reified the consequences of earlier policies in altered internationalconditions that further undermined Palestinian prospects. The collapse of theformer Soviet Union, followed within two years by the US-led invasion of Iraqto punish it for its aggression against Kuwait, reversed the former US support forthe Iraqi dictator (Chomsky 2000c: 53). Within a year, the weakened leadershipof the PLO agreed to the steps that led from Madrid to Oslo, Cairo, Wye, andother intervening stops in the “peace of the victors.”

For Palestinians under continued Israeli military occupation, the offers ofcantonization separated by a growing string of Israeli settlements, held out


276 Irene Gendzier

little promise. Peace under conditions of military rule in which human rights, letalone civil and political rights were violated, while access to water, electricityand food were threatened, was a form of pacification that assured continuedconflict. The “peace process,” Chomsky wrote, is about “whatever the UnitedStates happens to be doing, often blocking peace initiatives,” as the case maybe (1999b: 238). For those willing to confront the facts, however, it should “beunderstood as an impressive vindication of the rule of force in internationalaffairs, at both policy and doctrinal levels: the former, by virtue of its operativesignificance, the latter, in the light of the broad acceptance of the rejectioniststance that Washington had maintained in virtual isolation for many years”(1997: 160–1). The results can only be ominous, he warned, with “much painand suffering” ahead.2

After the Cold War

With the collapse of the Soviet Union and communist regimes of EasternEurope, the Cold War came to an end. What changed in terms of US foreignpolicy that had, for decades, been justified in the name of anti-communism?Heralded as the dawn of a new era, US officials promised an era of profit andprogress for all in the form of privatization and democratization. But, in practice,US policy endorsed the accelerated expansion of corporate power and profit inthe name of support for free trade and the dismantling of statist regimes. Theneoliberal orthodoxy of privatization and globalization led to deregulation ofcapital flow and systematic efforts to undermine labor and environmental pro-tection. These had little to do with free trade and even less with democracy,as Chomsky’s work demonstrated. They represented instead an accelerationof US monetary and trade policies redefined in the 1970s in an effort to stemthe competitive advantage of US major industrial allies, Japan and Germany.Thirty years later, in the absence of the Cold War, the new international orderinvolves an unheralded, if internationally contested, US military expansion thatis designed to protect the gains of the new world order, the latest phase of thatpersistent other war, the one between the global North and South.

Chomsky’s work exposed the pervasive mantras of the post-Cold War era,including the claims of free trade that masked the pursuit of protectionist mea-sures critical to the American economy such as the “huge state component ofthe economy, which undergirded all of high technology industry during the‘golden age of free market capitalism.’ ” He unmasked the “ideology of thedouble-edged ‘free market’ state protection and public subsidy for the rich,market discipline for the poor.” And in Rogue States, he confirmed that theindispensable accompaniment of this international order was the rule of force,as demonstrated by post-Cold War US policies in Panama, Colombia, Iraq,Israel and Palestine, and those of the US and its Western allies in East Timorand Kosovo. In A New Generation Draws the Line, Kosovo, East Timor and the



Standards of the West (2000d), Chomsky returned to a theme that dominated hisearliest writings on US policy in Vietnam, namely, the need to understand thenature of policy as a prerequisite to changing it. Unless one confronts the mythsat the center of foreign-policy justification, he argued, their implicit norms ofconduct “are likely to prevail if self-serving doctrine remains immune to crit-ical reflection, and moral truisms are kept at the margins of consciousness”(2000d: 21). The warning not only serves as an apt summary of fundamentalaspects of Chomsky’s political views, it represents the other side of globaliza-tion, with its widespread resistance and continued commitment to struggles forsocial justice and democracy.

The urgency of recovering the historical record was reinforced by theunprecedented nature of the events of September 11, 2001. No commentaryon Chomsky’s analysis of US foreign policy can ignore the depth of the ensu-ing crisis or the integrity of his response.

The attack on the World Trade Center in New York and the Pentagon inWashington left more than 3,000 dead (Talbot 2001: 16). But it was not the tollalone that rendered the event without precedent so much as what was perceivedas the nature of the target. For the first time since the War of 1812, propertiesand civilians within United States territory became the subjects of direct attack(Chomsky 2001b: 11).

In the US, a palpable sense of solidarity temporarily erased political andsocial boundaries in the face of what was perceived as a national emergency.The fear of further attacks and the risks of bio-terrorism intensified the gener-alized apprehension. Washington called for an international coalition againstterrorism. That, in practice, became a near-unilateral US and British war onAfghanistan, ostensibly in search of Usama bin Laden, the Saudi Muslimmilitant whose broad network of supporters were believed to be responsiblefor the September 11 attacks.

The vulnerability of the undisputed superpower shocked Americans, as itdid much of the world. Sympathy for Americans abroad was widespread. Itdid not, however, preclude increasing criticism of Washington’s preference forunilateral military action taken outside of existing international frameworkswith calamitous consequences for Afghan refugees and civilian casualties ofUS bombings.

Chomsky voiced these and a range of other pointed criticisms against theUS war in Afghanistan and the studied indifference to the toll of US policiesin the Middle East, the taboo subject forcibly opened by the terrible events ofSeptember 11. Against the increasing conformism of the mainstream US mediahe counseled resistance to intimidation and warned against those who “demandsilent obedience” (2001b: 118).

In characteristic fashion, Chomsky insisted that Americans confront the ter-rorist attacks as legacy of their government’s policies. In the case of Afghanistan,that meant coming to grips with past US support for the Taliban, no less than with


278 Irene Gendzier

the dark legacy of Washington’s new-found allies in that country, the forces ofthe Northern Alliance. He rejected claims that the US was under attack, insistingon the definition of the term in international law that undermined its applica-bility, while questioning Washington’s declaration of a “war against terrorism”in the light of its own indictment of international terrorism. Washington’s dec-laration was “a term of propaganda,” Chomsky argued. “There cannot be a waron terrorism led by the one state in the world that has been condemned forinternational terrorism and supported by major terrorist states like Russia andChina” (Chomsky & Hansen 2001: 7). Chomsky did not deny that the eventsof September 11 constituted terrorism, but “private terrorism” as opposed to“state terrorism” carried out by some of the above. And there was no mistakingChomsky’s view of what should be done in the wake of the “criminal atrocities”of September 11. Chomsky, like many other individuals in the US and abroad,argued in favor of “careful police work; a criminal investigation carried outby international authorities; the use of internationally sanctioned means, whichcould include force to apprehend the criminals; bring the criminals to justice;ensure that they have fair trials and international tribunals” (2001b: 91). It wasalso evident that the US had no intention of going before an international body,including the UN Security Council, to request authorization for its actions. Thereasons were not difficult to imagine. Chomsky reflected, “my speculation isthat the U.S. does not want to establish the principle that it has to defer to somehigher authority before carrying out the use of violence,” citing support for sucha position by Clinton and his Secretary of State (2001b: 4).

Within days of the events of September 11, Chomsky gave a succession oflive and written interviews in the US and abroad. He appeared in India andPakistan in late November, where he was described by Ejaz Haider of TheFriday Times of Pakistan as “the conscience that troubles everyone” (Haider:2001). His was a conscience that troubled those unprepared to question themeaning of US justification of the “war on terrorism,” or the risks of massstarvation facing some 7.5 million Afghan refugees, as international agenciesrepeatedly warned, or the scale of civilian casualties, however “unintended”as US officials claimed. More troubling to many – including leftwing intellec-tuals in the US – was Chomsky’s reminder that the scale of the September 11atrocity was not uncommon elsewhere in the world, including as a result of USpolicies such as the 1998 US attack on the pharmaceutical plant in the Sudan andthe catastrophic toll of the US-backed Contra war against Nicaragua, amongothers (2001b: 45–54, 56, 85–6). Not to confront such policies, Chomsky haslong argued and repeats anew, is to risk the escalation of violence “leading tostill further atrocities such as the one that is inciting the call for revenge,” hemaintained in interviews given within two weeks of September 11 (2001b: 26).

The extensive interview Chomsky gave to the online journal, Salon, notonly provided these and innumerable other insights into Chomsky’s views, it



offered an abridged guide to the moral foundations of his political beliefs. To theinterviewer who questioned what difference confrontation with past US policieswould make, he replied in favor of truth-telling as an “elementary step.” Thus,“if we were more honest about some of the things that we’ve done,” he claimed,“if we were honest, then we could at least evaluate what we do sanely. If we’redishonest, we know that whatever we do, only by the merest accident will itbe justified. The first elementary step is honesty. After that you can go on andconsider complicated issues on their merits” (Chomsky & Hansen 2001: 7). Onthe use of force, Chomsky conceded that he was no pacifist. “I think the useof force is sometimes legitimate” – but those who advocated it and particularlythe use of “extreme violence” have a heavy burden of proof to meet. “That,” heclaimed, was “a moral truism” (Chomsky and Hansen 2001: 9). Elsewhere in thecourse of the same interview, he argued that “if there is a principle that we applyto others, we must insist that the principle apply to us,” which translated into theworld of September 11 meant that as it was right to cry out against the “horribleatrocity” that had occurred in the US, it was no less so to denounce “a crimewhen we commit it against others.” As to what the US ought to do, Chomskyreplied unhesitatingly that it ought to “stop participating in atrocities,” citingexamples from East Timor to Kosovo to the Kurds in Turkey, among others.

The challenge was magnified by the US-led invasion of Iraq, a war that wasjustified by the G. W. Bush administration as the response to Iraq’s possessionof weapons of mass destruction that allegedly constituted an imminent threat tothe United States. In practice, the US-led war constituted a war of aggressionwhose justification was exposed as a form of mass deception. In fact, it was awar that intensified international terrorism and promoted the risks of weaponsproliferation at an incalculable cost in human suffering.

Against the intellectual apologists of deception, Chomsky repeatedlyaddressed the responsibility of intellectuals to denounce the crime of silenceand to address the veritable nature and consequences of policies carried outby the US and its allies in Iraq. He recognized the unprecedented internationaloutcry against the war. And, as did Bertrand Russell at the International WarCrimes Tribunal of 1967, Chomsky sought to – in Russell’s words – “arouseconsciousness in order to create mass resistance” (Chomsky 1971c: 1) to poli-cies deceptively cloaked in the name of liberation.

Chomsky’s voice was raised in support of free and rational inquiry that aimedat speaking truth and resisting power. It was a call, by “one of the truly great menof the twentieth century,” as he had earlier described Russell, that was heardaround the world by those with a common longing for survival and againsthegemony.


14 The responsibility of the intellectual

Jean Bricmont

The tone and unyielding criticism long ago landed him [Chomsky] in theSiberia of American discourse. It’s an undeserved fate. What Chomsky has tosay is legitimate. If there is anything new about our age, it is that Chomsky’squestions will eventually have to be answered. Agree with him or not, we loseout by not listening.

Business Week1

Introduction

When the South African freedom fighter Steve Biko said that “the most pow-erful tool in the hands of the oppressor is the mind of the oppressed,” he wasexpressing an idea very similar to that of David Hume who observed: “’Tistherefore, on opinion only that government is founded; and this maxim extendsto the most despotic and the most military governments, as well as to the mostfree and the most popular” (quoted in Chomsky 1991: 352). From the fall ofthe Shah of Iran to the changes in Eastern Europe or South Africa, examplesabound that illustrate this thesis: a powerful group may have at its disposal allthe military might that it wants, if the soldiers and officers, or even the vastmajority of the population, are not willing to follow orders, it is effectivelypowerless. Hence, the importance of the battle of ideas and the necessity forany ruling class who wants to maintain its privileges in the long run of “regi-menting the public mind every bit as much as an army regiments the bodies ofits soldiers.”2

The importance of thought control of the general population suggests pre-cisely that the role of the critical intellectual is crucial for any movement aimingat liberating social change. I shall argue that the writings of Noam Chomskyoffer an outstanding example of what a critical intellectual can do. Politicalactivities of (leftist) intellectuals often oscillate between two extremes: eitherthey absorb themselves entirely into militant work (usually when they areyoung) and do not really use their specific abilities as intellectuals; or theyretreat from that kind of involvement, but then limit themselves to expressing

280


The responsibility of the intellectual 281

moral indignation disconnected from genuine political analyses. Moral indig-nation is necessary, but it should be based on a rational analysis, rather than onemotional reactions, because our emotions are all too easily manipulated by thepower and the media.

I cannot possibly do justice to the thousands of pages that Chomsky has writ-ten exposing the lies of Western governments and the deceptions of apologeticintellectuals and journalists. I can only hope to induce readers to look at theevidence themselves and make their own judgments. I shall first explain somegeneral principles contrasting what intellectuals should do (next section) withwhat they actually do in Western societies as well as explaining the basic ide-ological mechanisms through which world affairs are usually presented (thirdsection). Once these are understood, it “only” takes hard work to show howthey function in various concrete situations, some of which3 will be discussedin the fourth section. Unlike many intellectuals who write about “ideology,”Chomsky puts facts and evidence before “theory.” However, there are somegeneral attitudes, as well as pitfalls to avoid, that can be gathered from hiswritings. The final section will be devoted to them.

What are intellectuals responsible for?

For Chomsky, the responsibility of the intellectual is to “tell the truth as bestone can, about things that matter, to the right audience” (1996a, b: 55). How-ever simple this statement may seem, every part of it requires some comment;first of all, the idea of “telling the truth” has become, for certain sectors ofthe Western intelligentsia that can loosely be described as “postmodernists,”highly problematic: truth is viewed as an illusion or as merely a tool of power.Chomsky has neither sympathy nor patience for such ideas.4 When asked aboutthe antagonistic relation between science and anarchism, he replies: “Withinthe anarchist tradition, there has been a certain feeling that there’s somethingregimented or oppressive about science itself . . . I’m totally out of sympa-thy with that attitude. There are no arguments that I know of for irrationality”(1987a: 22).

Chomsky views truth and rationality as weapons of the oppressed, not againstthem. Claims to truth, expertise, etc. are a different matter, of course, and areregularly used as a weapon by the strong against the weak; but Chomsky isopposed to what one might call the easy or lazy position, which is unfortunatelyquite widespread in the intellectual left nowadays, namely to reject all claimsto truth on spurious philosophical grounds (relativism, postmodernism). On thecontrary, the task of leftist intellectuals is to dig out the facts and correct abusivereasoning, but that of course presupposes that there is such a thing as a fact, thatarguments do matter, and that not all stories are equal. When Chomsky hears


282 Jean Bricmont

that one of the postmodernist’s complaints against “Reason” is that it tries toseparate the “real” and the “not real,” he replies:

At least, I know that I try to make this distinction, whether studying questions thatare hard, like the origins of human knowledge, or relatively easy, like the source andcharacter of U.S. foreign policy. In the latter case, for example, I would try, and urgeothers to try, to separate the real operative factors from the various tales that are spun inthe interests of power and privilege. If that is a fault, I plead guilty, and will compoundmy guilt by urging others to err in the same way.5

A second qualification concerns “things that matter.” Here, moral issues enter.One could write about the actions of Genghis Khan; that can be intellectuallyinteresting, no doubt, but since nothing can be done for his victims, the moralrelevance of such writings is nil. Turning to contemporary events, more delicatequestions arise; since nobody can deal with every possible issue, one shouldask which crimes to denounce and which lies to expose, at least primarily; inparticular, should one concentrate on those committed by one’s own country orby our enemies? The answer to that question, in the West, is straightforward, atleast when it concerns intellectuals in enemy states, say the Soviet Union in thepast. Their duty is to denounce, if they can, the crimes of their own state; whatthey say about other states does not matter much. In fact, if a Soviet intellectualdid denounce only Western crimes (committed e.g. by the United States orIsrael) he was regarded with contempt, because such denunciations, even ifthey were accurate, could only, within the context in which they took place,have the effect of strengthening the resolve and hence the violence of his ownstate. But when it comes to Western intellectuals, things change radically: ourduty is supposedly to denounce the crimes of our enemies, not primarily thoseof our own governments.6 But Chomsky refuses this double standard; for him“the crimes that matter” are those for which we are more directly responsibleand about which we can do something, namely, primarily, the actions of ourown governments.

Finally, there is the question of the audience. Chomsky strongly disagreeswith the Quaker slogan “speak truth to power” (1996a, b: 60–1), because theconstraints put by coercive institutions prevent people holding power in thoseinstitutions to act humanely. If the director of a major corporation suddenlybecame a human being concerned for the fate of the environment or for thewellbeing of his workers, the shareholders would immediately fire him, or hiscompany would go bankrupt. If the President of the United States suddenlytried to dismantle the CIA or the Pentagon, he would be impeached. That doesnot mean that people holding positions of power cannot become moral agents,but then they will dissociate themselves from those positions. Chomsky simplydoes not write for people holding positions of power; his writings are meant



primarily to be a service to activists. They can also be regarded as tools ofintellectual self-defense against the dominant discourse, the media etc.

How does ideology work?

Whole libraries of books have been written on the notion of ideology, fromthe point of view of Marxism, psychoanalysis, critical theory, etc. Comparedto those, Chomsky’s approach may seem terribly down-to-earth (but that mayturn out, upon reflection, to be an advantage rather than a defect). Anyway,ideology here will simply mean a mystification used by a social group in orderto maintain its power.

As a paradigmatic example of a self-serving mystification, although it is notone that Chomsky often uses, consider priesthoods in traditional societies. Bymaking people believe that they are the proper intermediaries between humansand some deities, and by convincing people that the current social arrangementsresult from the will of those deities, priests secure for themselves a privilegedposition under the benevolent umbrella of the ruler. In modern societies, priestsno longer have such an important ideological role as in the past, and their roleis played by what one might call the “secular priesthood,” namely, academics,intellectuals, journalists. The function of those experts, a function which is oftenwell rewarded, is to hide the way that private or state power actually works andto provide justifications, in terms of the pursuit of lofty ideals, for that powerand for the awful violence that its exercise often leads to.

The basic functioning of the dominant ideological discourse in the Westernsocieties is very easy to explain; when historians analyze a society in the distantpast, say the Roman empire or Genghis Khan, they will try to relate the actionsof the kings and rulers to their perceived economic and strategic interests.And, when they analyze the behavior of other actors, with less power, such aspriests or military men, they will do so in terms of the institutional constraintsimposed on them by the society under study. Such an attitude is taken forgranted without question; in particular, nobody feels that one has to explainthose behaviors on the assumption that the proclaimed intentions of the rulers aretheir true motivations. Quite the contrary, it is the “hidden” structure (politicalobjectives, or institutional constraints) discovered by the historian that will beused to analyze the official discourse of the rulers and explain them as somekind of ideological justification.

In fact, such an attitude extends to everyday life: nobody wants to buy aused car from an unknown seller who says first that the car is in excellentcondition and second that the price is so low because he/she is your friend. Andof course, this attitude also extends to the analysis of contemporary societies,such as the Soviet Union (before its collapse) or China. It would be hard to


284 Jean Bricmont

find a Western analyst who would try to understand the Soviet “intervention”in Czechoslovakia or Afghanistan mainly as an attempt to “defend socialism.”Rather, he/she will focus attention on the structure and rivalries within Sovietpower. There is, however, one exception to this general methodological attitude:namely, present Western societies. The attitude then becomes spontaneouslythe exact opposite: it is axiomatic that the motivations of our governments arepure and are basically identical to those presented in official discourse (defenseof human rights or democracy, etc). One is of course free to challenge theattainability of these objectives or the wisdom of those who try to implementthem; but to question the purity of the motivations or their legitimacy is to putoneself outside the bounds of respectable discourse.7

A standard argument offered to support the dominant view, at least implicitly,is that present Western societies are “really different,” both from past societies(including our own past) and from our barbaric enemies (such as Russia, China,the Arab world, etc.). That is because “we” (meaning our governments) aregenuinely concerned by democracy, human rights etc. Defenders of that thesiswill readily concede that we “erred in the past”; sure, colonialism was bad,slavery was abominable and during the Cold War, we supported far too manydictatorships. But now things have changed and we can at last all work togetherto build a better world.

Note that this thesis is often presented as an obvious truth because indeeddemocratic forms as well as human rights, which we are supposed to defendeverywhere, are generally better respected in Western countries than elsewhere.But that true observation implies nothing concerning the nature of the foreignpolicy of the West. It is perfectly possible, for example, that within a societywhere some democratic forms are respected, an economic oligarchy formsitself and obtains enough control over the government and the media to directthe foreign policy towards goals that are far from noble. But, even withoutadhering to that idea, one should agree that this “obvious truth” should beevaluated empirically, if possible. The way to test it, as is often the case whenone wants to test someone’s sincerity, is to consider pairs of situations, onewhere it would be easy to attain the proclaimed noble goals (defending humanrights for example) and one where it is hard. If one observes that, in the firstcase, nothing is done (or, worse, that many things are done that actually hurtthe proclaimed goals, e.g. supplying military aid to brutal regimes) while, inthe latter case, a great deal is done, for example sustaining major wars, but withlittle or no result, then the charge of hypocrisy cannot be avoided. A good deal ofChomsky’s work is devoted precisely to the analysis of many pairs of examplesthat support this charge, both concerning the actions of Western governmentsand the way the mainstream media present them.8

A second structural feature of the way ideology works in the West is to haveas strong a debate as possible within limited bounds. The typical form that such



a debate takes is to concentrate on the means to achieve some – by assumptionnoble – goal and to avoid carefully any question regarding the sincerity of theenterprise. Thus, the debate will typically be centered on questions such as:do we have enough power, resolve, etc. to achieve our goals? Are our leaderssufficiently clever, determined, and strong? The more vociferous the debates,the more the implicit assumptions concerning the nobility of the intentions cometo be reinforced. The Vietnam War is a good example: the hawks maintain tothis day that with more determination, the United States could have “savedVietnam” and that it was “lost,” because the war effort was stabbed in the backby the peace movement. To that, the doves reply that the effort was doomedto fail, that it was too costly to the United States and maybe even killed toomany people. At the end of the war, the New York Times gave a good summaryof the terms of that debate: “There are those Americans who believe that thewar to preserve a non-Communist, independent South Vietnam could have beenwaged differently. There are other Americans who believe that a viable, non-Communist South Vietnam was always a myth . . . A decade of fierce polemicshas failed to resolve this ongoing quarrel.”

To this Chomsky, together with Ed Herman, replied in an (of course, unpub-lished) letter, that “there was a third position: That apart from its prospectsfor success, the United States has neither the authority nor the competence tointervene in the internal affairs of Vietnam. This was the position of much of theauthentic peace movement, that is, those who opposed the war because it waswrong, not merely because it was unsuccessful” (1977c: 37). The Vietnam Warwill be discussed in the next section, but just to see how odd the summary of theNew York Times is, compare it with the following statement: “There are thoseSoviets who believe that the war to preserve a non-fundamentalist, independentAfghanistan could have been waged differently. There are other Soviets whobelieve that a viable, non-fundamentalist Afghanistan was always a myth . . .A decade of fierce polemics has failed to resolve this ongoing quarrel.” In thatsituation, nobody has any difficulty seeing that there is a third position, the onethat Chomsky and Herman adopt for the Vietnam War; but, here, the third posi-tion is hidden by the raging debate between the two “admissible” ones. Let usnow turn to concrete examples of how all these ideological mechanisms work.

Examples

The Indochina Wars

Chomsky always considered the Vietnam War to be part of a larger framework,which one might call the one of the “rotten apple.”9 At the end of World War II,the leaders of the United States saw that their country could become the solesuperpower and worked to obtain and maintain that position, which led to


286 Jean Bricmont

overt and covert interventions in Greece, Guatemala, the Dominican Republic,Brazil, Chile, Lebanon among others.10 The idea of the rotten apple is thatany country, no matter how small, that can escape from United States globaldominion and develop itself independently or, even worse, set as one of itspriorities the wellbeing of its own population, is a danger, because it mightprovide the wrong model for others (hence, spreading the “rot”). In fact, thesmaller and the poorer the country (Cuba, Vietnam, or Grenada), the greaterthe threat, because the more impressive such an example would be.

The movement of national liberation in Vietnam did pose such a threat of“independence” after 1945. This led the United States not only to finance theFrench war in Indochina, but to undermine the Geneva 1954 peace agreementsthat ended that part of the war; it then supported a terrorist war by the Saigonregime against its own population, sent the US air force to bomb the countrysideas early as 1962 (long before any North Vietnamese soldiers appeared in theSouth), and drove millions of people into camps called “strategic hamlets,” inorder to cut the South Vietnamese guerillas from their popular base. This esca-lated into a full-scale intervention with hundreds of thousands of ground troopsand massive bombing of North Vietnam (and even more of South Vietnam) in1965. Eventually, the war proved too costly for the US elites and they started to“oppose the war,” but on grounds of effectiveness, not on any basis of principle.Chomsky is quite sarcastic with respect to that kind of opposition, which hedoes not view as morally superior to the possible opposition of German generalsto their own war after Stalingrad.

Chomsky’s criticism of “the new mandarins”11 was directed primarily atintellectuals who supported United States power, notably in its aggressive warsagainst recalcitrant small countries such as Vietnam and, later, Nicaragua. How-ever, his criticisms of various forms of misguided or excessively soft oppositionto the war more clearly sets him apart from the left-liberal mainstream. Indeed,it is precisely by analyzing the discourse of opponents to the war and someof their delusions that one can appreciate how much even opponents can beindoctrinated by the dominant ideology.

Chomsky illustrates the power of our propaganda system compared withthat of the Soviets by citing the story of Danchev, a courageous newscasterin Moscow who denounced the Russian invasion of Afghanistan (Chomsky1987a: 223–6). Danchev actually called upon the rebels to resist. He was sentto a psychiatric hospital by the Soviets (but returned later to his position). Hisactions were praised in the West.

Notice that Danchev saw what anyone outside the Soviet Union could see,that the Soviets had invaded Afghanistan. This was contrary to official Sovietdoctrine, which maintained that the Soviet Union was defending Afghanistanagainst CIA-supported rebels based in Pakistan and that they were invited inby the Afghan government. That defense was dismissed in the West on the



grounds that the government was illegitimate, a mere Soviet puppet. Chomskypoints out, however, that the same Westerners who saw the Soviet action as aninvasion and praised Danchev for not only seeing it but denouncing it should, ifthey had been consistent in their moral evaluations and their praise, have beendoing exactly what Danchev did, only directing their denunciations against theUS government. For, if anything, it was even clearer that the United Stateshad invaded Vietnam – that the US wanted to establish hegemony in SoutheastAsia and that the South Vietnam government that “invited” them was a pup-pet. And the responsibility of the Western intellectual was greater, too, forDanchev acted in spite of the threat of prison or a psychiatric hospital, wherethe Westerner was subject to no such threat. Nevertheless, the expression “USinvasion of South Vietnam” was virtually never heard, even in the peace move-ment. And only a very few marginalized critics recognized this invasion forwhat it was.

In the Soviet system, it was not thought itself that was controlled (even if thatwas the goal), but expressions of thought. Our system of “brainwashing underfreedom,” which does not control expressions of thought directly, and in factencourages debates, as long as the purity of the intentions of our leaders remainsunchallenged, is in fact a much more effective system of thought control thanthe Soviet one.

As another example of the kind of illusions that are widespread within theleft, Chomsky criticizes misplaced notions of scientific and moral expertise.During the Vietnam War, in 1965, a conference was organized that tried toarticulate the insights of social scientists and other individuals from “varioustheological, philosophical and humanist traditions” in order to “find solutionsthat are more consistent with fundamental human values than current Americanpolicy in Vietnam has turned out to be.” Chomsky’s reaction was biting:

The only debatable issue, it seems to me, is whether it is more ridiculous to turn toexperts in social theory for general well-confirmed propositions, or to specialists in thegreat religions and philosophical systems for insights into fundamental human values . . .if there is a body of theory, well tested and verified, that applies to the conduct of foreignaffairs or the resolution of domestic or international conflicts, its existence has been kepta well-guarded secret. (1987a: 71–2)

Of course, Chomsky does not deny that one should seek from the social scienceswhatever information can be obtained. But he is violently opposed to the cult ofthe expert, “both self-serving and fraudulent.” Many intellectuals love to playthe role of “advisors to the Prince” and to claim to be technical experts who will,using some deep principles of social theory that are inaccessible to ordinarypeople, propose solutions to current problems. But if those problems, like theVietnam War and its conduct, are basically due to relationships of power anddomination, as Chomsky claims, then the people who present them as merely


288 Jean Bricmont

technical problems only help those who hold a position of power to hide theiractions behind a discourse based on socially neutral categories.12

Finally, another illusion that Chomsky often denounces is the idea that theUnited States lost the Vietnam War.13 This is only partly true: the United Stateshad a maximal objective, which was to impose there their favorite regime, andthat failed. But the minimal goal, which was to prevent the rot from spreading,by destroying any hope of autonomous development in Indochina, was com-pletely successful. First of all, a heavily bombarded Vietnam was not exactlyan attractive option for its neighbors but, also, the isolation imposed on a devas-tated Indochina after the war eventually led to its reintegration into the dominantworld system. Moreover, the United States was able, during the Vietnam War, toconsolidate its position in neighboring countries such as Thailand and Indonesia(through murderous coups, such as the one of Suharto in Indonesia in 1965,where, as former CIA agent Raph McGehee puts it, the agency created “condi-tions that led to the massacre of at least half a million Indonesians” [Chomsky1987a: 305]).

The reconstruction of the imperial ideology

Nevertheless, the image of the United States was tarnished and its populationhad become massively opposed to direct foreign interventions. It was thereforenecessary for the US elite to continue the war at the ideological level, and to tryto construct new justifications for future interventions that were to take placein the 1980s and the 1990s. Notable examples include Central America andIraq, leading to the deaths of hundreds of thousands through civil wars, coun-terinsurgency campaigns, and embargoes. It is interesting to note Chomsky’sreactions, and to contrast them with the evolution of the majority of the intellec-tuals who have been opposed to the war, and who ended up basically acceptingthe mainstream discourse.

The atrocities committed by the Khmer Rouge as well as the exodus of the“boat people” fleeing Vietnam were to the mainstream media and the liberalintelligentsia heaven-sent gifts. The atrocities and the exodus could be used toshame the anti-war movement (“see, we told you what would happen”) andto lay the ground for future interventions. Some people who opposed the warfelt, for example, that they had a “special responsibility” to denounce humanrights violations in Vietnam; some even came to reconsider their opposition tothe war.14 But, for Chomsky, this was an instance where people fell victim ofthe assumptions of the ideological system. To understand why, consider Sovietdissidents who would have opposed the invasion of Afghanistan. The fact thatthe fundamentalist Islamic leader Hekmatyar committed atrocious crimes afterthe Soviet withdrawal in no way implies that those dissidents have a “specialresponsibility” to denounce them. Those dissidents may very well say that the



Soviet intervention led to a disaster, that, by opposing that intervention, theydid all they could to prevent it and that the further consequences of that disastershould be faulted on those who carried on and supported the intervention, noton those who opposed it. But the Western analogue of this straightforwardargument is almost never heard.

In fact, already in 1970, Chomsky anticipated what was going to happen and,after quoting Kant’s remarks on the French Revolution: “one must be free tolearn how to make use of one’s powers freely and usefully. The first attemptswill surely be brutal and will lead to a state of affairs more painful and dangerousthan the former condition under the dominance but also the protection of anexternal authority,” he adds:

no rational person will approve of violence and terror. In particular, the terror of thepostrevolutionary state, fallen into the hands of a grim autocracy, has more than oncereached indescribable levels of savagery. Yet no person of understanding or humanitywill too quickly condemn the violence that often occurs when long-subdued masses taketheir first steps toward liberty and social reconstruction.15

Moreover, the criticism by American war resisters of human rights violationsor other indefensible acts committed by the victims of American violence againraises the question of the right audience. Given that any criticism of the victimsof the US assault, made in the US press, would very likely be used to further“bleed Vietnam,” through embargoes and other sanctions,16 its only likely resultwould be to contribute to increased human suffering. However, unlike someleftists, Chomsky does not rule out such criticisms – in fact, he is himself oftenvery critical of revolutionary societies such as Cuba or Nicaragua17 – but heinsists that those criticisms should be put into proper perspective and that weshould not ignore the actual human impact of our declarations. This attitude isquite distinct from the moral posturing with which all too many intellectualsseek to display their ideological purity.18

A related issue that aroused considerable fury among Chomsky’s opponentswas his alleged “defense of Pol Pot.” Of course, no such defense ever occurred.19

The main theme of Chomsky & Herman (1979a, b, 1988) was a comparisonbetween the responses of the media to the Khmer Rouge massacres in Cambodiaand to the Indonesian massacres in East Timor at the same time and similar inscale (relative to the population). In the first case, where the massacres could beblamed on an official enemy, virtuous indignation and gross exaggerations werethe rule; in the case of a massacre committed by a friendly state (Indonesia),almost total silence. The issue for Chomsky and Herman was not the nature ofthe Khmer Rouge regime, which was certainly bad enough, but the propagandistattitude toward an official enemy, where the only rule is “anything goes.” Amechanism much like religious bigotry is set in motion: everybody feels theneed to show that he is more virtuous than his neighbor in denouncing the crimes


290 Jean Bricmont

of the enemy; facts need not be checked and the grossest lies are consideredto prove only the sanctity of their authors. Similar mechanisms are at worknowadays concerning Iran or North Korea. Whoever objects, even on perfectlydemonstrable factual grounds, is regarded as a heretic and treated as such.Inasmuch as it was quite clear to everybody that there was nothing that could bedone in Cambodia at that time, the discourse about the Khmer Rouge was purelyideological. Indeed, the United States could hardly have intervened in a countryfrom which they had just withdrawn, nor did anyone else offer to intervene.Moreover, when the Vietnamese eventually did intervene and terminated thePol Pot regime, the utter hypocrisy of the Western powers became obvious,since they strongly opposed that intervention, continuing to recognize the PolPot regime at the United Nations and supporting it militarily in indirect ways(Pilger 1998).

By way of comparison, consider the story of the Belgian babies killed by theGermans during World War I, a classic story of disinformation. While there isno doubt that horrible atrocities were committed by the German troops whenthey invaded Belgium in 1914, there is also no doubt that those atrocities weregrossly exaggerated by the French and British propaganda machines, especiallyin order to draw the United States into the war. Therefore, pacifists, like BertrandRussell, had to put those atrocities into proper perspective, to get the factsstraight and in general to do everything they could to prevent the militaristsfrom using real or imagined atrocities to stir up emotions, thereby encouragingthe general population to accept, through the continuation of the war, furtherkillings, and on a far greater scale (killings that, of course, did happen; but thisdoes not mean that the efforts of the pacifists were misguided).20

Human rights: rhetoric and reality

The universality of the 1948 Universal Declaration of Human Rights is anotherobject lesson in the hypocrisy of Western governments and mainstream intel-lectuals. The standard view about it in the West is that there are some “badguys” (mostly African or Asian leaders and relativist intellectuals at home)who invoke cultural differences in order to deny the universal value of that dec-laration. Chomsky takes the declaration and its universality very seriously. Buthe does not agree with the standard view about who are the real “relativists.”21

Consider first the Civil and Political Rights, e.g. article 14 that grants the rightto seek abroad asylum from persecution. Its implementation is extraordinarilypoliticized by the United States: of the more that 24,000 Haitians interceptedby US forces from 1981 through 1990, 11 were granted asylum, in comparisonwith 75,000 out of 75,000 Cubans. Or consider article 13, granting “the right toleave any country, including his own.” This was constantly invoked with greatpassion against the refusal by the Soviet Union to allow Jews to leave. But the



end of that article, which says “and to return to his country” is ignored. Tounderstand why, note that a day after the Universal Declaration was ratified, theUnited Nations passed Resolution 194, which affirms the right of Palestiniansto return to their homes (or to receive compensations); note also that this right,although regularly affirmed since 1948, is not even taken into account by theso-called Oslo peace process between Israel and the Palestinians.

The declaration also contains Economic, Social and Cultural Rights, includ-ing a right to health care, social security and adequate standard of living(article 25). Whatever one thinks of those rights, they are part of the decla-ration and are as binding on the signatories as any other part of the UniversalDeclaration. Nevertheless, the United States ambassador to the United Nations,Jeane Kirkpatrick, could call them a “letter to Santa Claus” without provokingmuch reaction. The notion of an adequate standard of living may be somewhatvague, but the fact that millions of people suffer from hunger in such an incred-ibly rich country as the United States is a clear violation of article 25. A similarremark applies to the three-fold increase in child poverty in Thatcher’s Britain.

In the West, the Civil Rights part of the Universal Declaration is held tohave absolute priority over the Economic and Social parts. (This is in itself aninteresting example of “relativism” – just think of the reactions in the West ifsome Third World leader called the first part a “letter to Santa Claus”). In casethis strikes you as obvious, imagine yourself being one of those two or threebillion people (about half of mankind) that have to survive on more or less twodollars a day. What would be your view of the economic and social policiesof the Cuban government, assuming that all relevant facts were available toyou (of course, they are not available to the wretched of the earth, and forobvious reasons)? How would you weigh Cuban efforts to maintain publichealth, education and availability of basic necessities for the poor versus thelimitations imposed on civil liberties? Consider that these efforts continuedlong after Cuba had stopped being “subsidized” by the Soviet Union22 – andwhile it suffers from a very severe embargo as well as from numerous acts ofsabotage caused by the single superpower, which forces them to divert resourcesto defense, counter-spying etc. Although Chomsky would certainly not usesocio-economic rights to justify the abandonment of civil liberties, the contrastbetween the unequal emphases put on the two sorts of rights shows that theissue of “relativism” is not quite as simple as it is portrayed in the dominantWestern discourse.

A critical attitude with respect to the dominant discourse of human rightsoften provokes indignant hysteria among Western intellectuals. The only plau-sible explanation is that these intellectuals feel threatened – not because ofany use of power against them, but because they are being shown to be self-righteous hypocrites in the willing service of power, not genuine seekers oftruth. They enjoy the privilege of free expression, and do not exercise it. Their


292 Jean Bricmont

form of hypocrisy, regrettably, only paves the way towards more atrocities andkillings in US client states, where critical intellectuals really are threatened.The following comments, taken from the Jesuit Salvadorian journal Processo,illustrate this clearly:

If Lech Walesa had been doing his organizing work in El Salvador, he would havealready entered into the ranks of the disappeared – at the hand of “heavily armed mendressed in civilian clothes”; or have been blown to pieces in a dynamite attack on hisunion headquarters. If Alexander Dubcek were a politician in our country, he wouldhave been assassinated like Hector Oquelı [the social democratic leader assassinated inGuatemala, by Salvadorian death squads, according to the Guatemalan government]. IfAndrei Sakharov had worked here in favour of human rights, he would have met the samefate as Herbert Anaya [one of the many murdered leaders of the independent SalvadorianHuman Rights Commission CDHES]. If Ota-Sik or Vaclav Havel had been carrying outtheir intellectual work in El Salvador, they would have woken up one sinister morning,lying on the patio of a university campus with their heads destroyed by the bullets of anelite army battalion. (Chomsky 1991: 354–5)

To summarize, the aspirations embodied in the Universal Declaration couldbe a powerful tool to challenge the powers that be. However, thanks to a veryselective reading of the declaration and to a careful choice of targets of indigna-tion, mainstream Western intellectuals have managed to make this declarationalmost entirely harmless for Western governments and, in fact, have allowedthem to use it cynically for their own purposes.

Hopes and challenges for the future

What is to be done? Unlike many other intellectuals, Chomsky is quite carefulwhen he tries to answer that question. First, consistent with his libertarian ideals,he feels that each individual has to make his own choices. Besides, he is opposedto the attitude, rather widespread among intellectuals, that consists in offering(theoretical) solutions to complicated situations, like the Balkans, in which theyare not directly implicated. Finally, since he regards a well-tested science thatwould give answers to difficult social problems as a “well-kept secret,” he isvery critical of various ideologies that masquerade as sciences and that claim togive all-encompassing solutions to human problems. For him, progress in thebuilding of a humane society has to be based on experimentation and struggleand there is no point in treating either Marxism, or the free market or evenanarchism as a religious dogma.

In politics, Chomsky teaches by means of example, and is often at his bestin adverse situations. There is little doubt that, during and after the collapseof the Soviet system, the world has gone through one of the most reactionaryperiods of its history, similar to the restoration of monarchic order throughout



Europe after Waterloo, but now taking place on a world scale. This generalized“rollback” of the gains of workers in the rich countries as well as of the advancesmade by developing nations, was made possible, in part, by the discouragementthat affected a good part of the left. It is interesting to see how Chomsky reactedto that situation.

First, the following excerpt, where Chomsky quotes excerpts from the workof Bakunin, clearly shows the differences between him and the dominant trendswithin Marxism:

“The organization and the rule of society by socialist savants,” [Bakunin] wrote, “is theworst of all despotic governments.” The leaders of the Communist party will proceed “toliberate [the people] in their own way,” concentrating “all administrative power in theirown strong hands, because the ignorant people are in need of a strong guardianship . . .[the mass of the people will be] under the direct command of the state engineers, whowill constitute the new privileged political-scientific class.” For the proletariat, the newregime “will, in reality, be nothing but a barracks” under the control of a Red bureaucracy.But surely it is “heresy against common sense and historical experience” to believe that “agroup of individuals, even the most intelligent and best-intentioned, would be capableof becoming the mind, the soul, the directing and unifying will of the revolutionarymovement and the economic organization of the proletariat of all lands.” In fact, the“learned minority, which presumes to express the will of the people,” will rule in “apseudo-representative government” that will “serve to conceal the domination of themasses by a handful of privileged elite.”23

For Chomsky, the Soviet Union was an example of rapid industrializationof an underdeveloped country, but was very far from anything that one mightcall “socialism.” And so, having had no illusions to lose about the nature ofSoviet “socialism,” Chomsky does not join the chorus of those who celebrate(or lament) the “failure of socialism.” He also notes that some people, whoalways claimed to be anti-Leninist leftists but felt “discouraged” after 1991,“were more deeply committed to Leninism than they believed” (1996c).

But Chomsky is not discouraged. Among reasons for hope, one should stressthat during the second half of the twentieth century, remarkable developmentsalong the path of human liberation have taken place: the liberation of the ThirdWorld from the most oppressive forms of Western domination, the women’s lib-eration movement, enormous progress in the struggle against racism, changesin education both in the family and at school, sexual liberation; despite impres-sions to the contrary, the years between 1945 and 1980 may have been amongthe most emancipating in human history (1980 being roughly the beginning ofthe “rollback” period). Moreover, as Chomsky often stresses, the part of the anti-Vietnam War movement that was based on a principled opposition to the warrepresented a major breakthrough in the intellectual and moral level of theAmerican youth (probably one of the reasons why so much effort is devoted tomisrepresenting and slandering that period of history). The current worldwide


294 Jean Bricmont

opposition to corporate controlled “globalization,” also waged in large part bythe youth, shows an even greater degree of social awareness.24

However, if there is a sobering thought that one may have about the progressof civilization, it is that, at the same time as our proclaimed values have becomemore truly universal, there has been a considerable rise in hypocrisy. A leaderwho bases his rule mostly on, say, a racist ideology does not need to be toohypocritical. But if a violent power such as the United States constantly invokesChristian charity, international law or human rights, it needs the help of anintellectual class who will distort the facts and spread illusions in order tomaintain its rule. It is unfortunate that so many intellectuals are quite willingto play that role; but people like Chomsky also remind us that this choice isnot inevitable and that intellectuals can, if they want, play an important rolein the construction of a truly humane society. Indeed, the struggle to exposethe hypocrisy of the rulers may very well be an important element in the nextemancipatory stage of mankind, during which, amidst and against the presenthorrors of capitalism, a radical post-Marxist left will arise, one that will reclaimthe values of the Enlightenment and hopefully make them realize their fullpotential.


Notes

INTRODUCTION

1. Some thinkers, usually classified as Romantics, can be included on the list because –unlike most other Romantics – they recognize that innate (native) conceptualresources are needed to exercise the kind of freedom and creativity that is charac-teristic of romanticism. Relevant individuals include Wilhelm von Humboldt, A. W.Schlegel, and Samuel Taylor Coleridge (Chomsky 1966, 2002a).

2. Traditional empiricists were not much concerned with reference, perhaps becausethey, like the traditional rationalists, were more sensitive to skeptical arguments aboutmind–world relationships than today’s “realist”-inclined group.

3. For Putnam (1975), a linguistic meaning is a vector that includes a word’s supposedscientific object as a part of its meaning. For Fodor (1998), who in some respects isa rationalist but in his externalism is not, a word’s meaning is solely its (external, heinsists) denotational target. As Fodor’s presence in the externalist camp suggests, it isunclear how this form of concept/meaning externalism relates to empiricist learningviews. Putnam is a learning empiricist (Putnam 1967), but his defense of meaningexternalism does not speak explicitly to learning. Kripke and Burge do not committhemselves, although their Wittgensteinian views of language suggest sympathy forthe training picture.

4. Columnist Jonathan Kay (2004), writing in Canada’s National Post magazineBusiness, follows Milton Freedman in arguing that it is fine for corporations to per-form socially responsible acts (selling a limited-supply drug at a price below what itcould fetch from the well off, for example) that improve their public image and thusaccord them more profit in the long run. But their agents (boards of directors) shouldensure that their actions are “amoral”: they must not see their task as serving the pub-lic interest unless appearing to do so increases profit. In influencing public officialsto serve corporate needs (profit), then, corporations are only doing what serves theirinterests, not human interests. It is misleading, of course, to think of corporations –or any institutions – as natural objects, or of their “rights” as natural (cp. Burchill &Chomsky 1998: 19). Corporations are artefacts; they are made to serve the interestsof the rich. Their “rights,” unlike those of humans, are not natural.

1 CHOMSKY’S SCIENCE OF LANGUAGE

I am grateful to Annabel Cormack and Jim McGilvray for comments on an earlierversion of this chapter.

295


296 Notes to pages 22–48

1. Hence, requiring that linguistic rules be accessible to consciousness, for instance(Quine 1972: 442), is unmotivated.

2. Chomsky (2000a). Chomsky’s insights into the human condition in the politicaldomain are akin to Dostoevsky’s, in the linguistic domain to Darwin’s.

3. This locution is used regularly by Chomsky (1995c: 2, 2000a: 1) to refer to whatis “in our minds, ultimately in our brains” (1980: 5). He repeatedly dismisses theso-called “mind-body problem” as unformulable in the absence of a coherent notionof body (1988b, 1995a).

4. Note that “generative” is used with systematic ambiguity in the literature. First,it appears as a near synonym of “explicit,” on which reading almost all currentgrammars and theories are “generative”; second, it is used to refer specificallyto work in (transformational) generative grammar associated over the years withChomsky and his followers.

5. Chomsky (e.g. 2000a: 6) often attributes this insight to Wilhelm von Humboldt(1767–1835).

6. For introductory discussion, see Fromkin (1996); for Chomsky’s view, see his(1999a).

7. Transformations change one structure into another structure, relating (for instance)statements to questions. They were central to Chomsky’s work in the 1950s (seeespecially 1955, 1957a), and gave their name to the theory. They have been lessvisible in more recent incarnations of the theory, but remain the locus of research(and disagreement) by Chomsky and those influenced by him.

8. For Merge and Move, see Chomsky (1995c).9. As determined by Binding Theory (Chomsky 1995c: 92ff).

10. For selection, see Chomsky (1986).11. See Wiggins (1997) for a contrary view.12. In fact, perfection in the sense of providing an optimal solution to particular problems

is pervasive in biology: e.g. the existence of universal scaling laws (West et al. 2000).13. For discussion of the differences between Chomsky’s and Fodor’s notions of mod-

ularity, see Smith (2003).14. This claim is, of course, contested – e.g. Tomasello (2000).15. The question was formulated as Plato’s Problem in Chomsky (1986); Principles and

Parameters theory is first set out explicitly in Chomsky (1981a,b).

2 PLATO’S PROBLEM, UG, AND THE LANGUAGE ORGAN

This paper incorporates material from Lightfoot (1999) and from Anderson &Lightfoot (2000), and I remain indebted to the people thanked there. Now Ithank Brigitte Bedos-Rezak for discussing an earlier draft from the viewpoint of anon-linguist.

1. This represents a different analysis from the ones given in Lightfoot (1999) andAnderson & Lightfoot (2000). The silent, understood element following the secondis might be some form of the adjective happy or, following the analysis of Chomsky(1977c), the trace of a WH element.

2. The technical notion of “c-command” is relevant here. In (5b), the NP Jay’s brotherc-commands him, but the NP that consists of Jay, contained within the larger NP,does not c-command him and therefore is unavailable for indexing.


Notes to pages 50–71 297

3. This reflects the critical period hypothesis, that there is a genetically determinedwindow of opportunity for language acquisition. If a child fails to develop languageduring this period, for whatever reason, then she will never attain a normal grammarand native-like competence. For good discussion, see Smith (1999: 120ff.).

4. SLI represents a genetically determined condition in which the language capacityis impaired in precise ways, while other mental domains seem to be normal. Theproblems relate to plurals, tense, gender, and aspect, but not temporal adverbs; theimpairment seems to affect specifically those parts of grammar where abstract mor-phological features are implicated. There is some controversy about whether thecases grouped together under this diagnosis should in fact be taken together, but thedistribution of SLI in some well-studied populations has been shown (in both epi-demiological and genetic studies [Tomblin 1997] to be that of a simple Mendeliantrait (Gopnik 1990; Gopnik & Crago 1991), perhaps even with a specific, identifiablechromosomal location.

3 GRAMMAR, LEVELS, AND BIOLOGY

I would like to acknowledge the very helpful discussion and/or suggestions of CedricBoeckx, Noam Chomsky, and, especially, Jim McGilvray. Much of this chapter waswritten while I was a Fellow at the Center for Advanced Study in the BehavioralSciences. I am grateful for the support provided by the John D. and Catherine T.MacArthur Foundation, grant no. 95-32005-0.

1. Although Chomsky’s use of the term “generative” has remained constant throughouthis writings, in the field at large the term has sometimes come to be associated withwhat Chomsky calls the creative aspect of language use – the ability of speakersto produce brand new sentences appropriate to situations, but in no sense directlytriggered by them (in the stimulus-response) sense.

2. Chomsky and Halle (1968) argue for a considerably more abstract representation,but that difference need not concern us at the moment.

3. Words that are not subject to the alternation (such as fife, with plural fifes) simplyhave the element f at the morphophonemic level instead of F.

4. To allow for the generation of these examples, we must add to the phrase struc-ture grammar in (13) either the possibility of rewriting NP as N with no determiner(NP → N) or the possibility of a silent Det. The choice between these two possi-bilities raises interesting questions, but is not relevant to the issue being discussedhere.

5. Note that in this model, for a complex sentence there is no one initial P-markerthat could provide the basis for semantic interpretation, since each clause has itsown separate P-marker. Further, the final derived P-marker also does not providesufficient basis for semantic interpretation, since quite often items occur displacedfrom the positions where they are interpreted, as in passive sentences like Marywas selected, where Mary is the understood (and underlying) object of the verbyet occurs in surface subject position. Thus, the T-marker is crucial to semanticinterpretation.

6. Here, as in much work in the first couple of decades of generative syntax, research inphonology had led the way. In phonology, the concept of cyclic application of rulesgoes back at least to Chomsky et al. (1956).



7. Below, we will see that Chomsky, in developing his recent minimalist program, putsforward an argument rather reminiscent of this one.

8. One of the standard arguments for traces (though not one discussed by Chomsky[1973a]) comes from a contraction process in colloquial English by which want andto become wanna when they are immediately adjacent:

(i) a. You want to solve this problemb. You wanna solve this problem

That this process demands adjacency between want and to is shown by the impos-sibility of (iib), based on (iia).

(ii) a. You want someone to solve this problemb. *You wanna someone solve this problem

The relevant (and surprising) property of wanna contraction is that even if someonein (ii) is replaced by the corresponding interrogative who, which is then displacedby movement, contraction is still blocked:

(iii) a. Who do you want to solve this problem?b. *Who do you wanna solve this problem?

Superficially, there does not seem to be anything intervening between want and toin (iiia), hence, nothing to prevent contraction. But if we assume that movementleaves a trace, then there is, in fact, something intervening – the trace of Who.

9. Government and Binding are two important technical terms of the theory. Chomskycame to feel that “Government–Binding theory” was a misleading appellation, sincethere were many other equally important technical terms in the theory, so thatthese two should not be singled out. Chomsky’s preferred term was “Principles andParameters theory,” a name highlighting the fact that at the center of the frameworkwere linguistic universals (principles) and simple limited ways that languages candiffer (parameters).

10. Chomsky felt that the term “deep structure” conveyed the false impression thatthat level was what was important or profound about language, as opposed to thesuperficial or trivial surface structure. In reality, the levels played equally importanttechnical roles in the theory, and “surface” properties, e.g. all of phonology, aresurely as significant and rich in structure as any others.

11. (45) is a version of the third (“Condition C”) of three conditions on anaphoraproposed in Chomsky (1981a).

12. Chomsky puts it this way: “there are no levels of linguistic structure apart from thetwo interface levels PF and LF; specifically no levels of D-Structure or S-Structure”(1995c: 219). The apparent qualification is important, because none of Chomsky’sminimalist writings addresses the early arguments for the levels discussed abovedirectly relevant to phonology, such as morphophonemics and phonemics. Thoselevels evidently remain.

13. “Pied-piping” is the felicitous term introduced by Ross (1967) for situations wherean item is compelled to move, but additional material follows along with it.

14. See also Jackendoff (1972), Lasnik (1972, 1976).15. See Chomsky (1999a, 2000b), the latter of which was actually written earlier.



4 HOW THE BRAIN BEGETS LANGUAGE

I am sincerely grateful to the deaf and hearing parents and their children whogave me and my students and assistants their time, trust, and good humor. Ialso thank Kevin Dunbar, and my students and research assistants who helpedin aspects of the analyses presented in this chapter, including Siobhan Holowka,Ioulia Kovelman, Gissella Santayana, Ulana Harasymowycz, Kristine Gauna andElizabeth Norton. All research was funded by the following grants to L. A.Petitto: Social Sciences and Humanities Research Council of Canada, NaturalSciences and Engineering Research Council of Canada, the Spencer Foundation,USA, and a Dartmouth College research grant. I also thank the GuggenheimFoundation for a generous fellowship. Address all correspondence to Laura AnnPetitto, Department of Psychological & Brain Sciences and Department of Edu-cation, Dartmouth College, Hanover, New Hampshire, 03755, USA. For moreinformation see Petitto Research Web Site: http://www.dartmouth.edu/∼lpetitto orPetitto Laboratory Web site: http://www.dartmouth.edu/∼lpetitto/lab. Email: [email protected].

1. For videotape examples of Optotrak babies producing linguistic versus non-linguistichand activity see http://www.dartmouth.edu/∼lpetitto and click on “Nature Supple-ment Petitto et al. 2001” (top upper right).

2. For videotape examples of baby linguistic versus non-linguistic, as well as smiling–mouth asymmetries, see http://www.dartmouth.edu/∼lpetitto and click on “Science2002 supplement.”

3. For videotape examples of lexical and phonetic-syllabic hand stimuli in ASL seehttp://www.dartmouth.edu/∼lpetitto and click on “PNAS supplement, Petitto et al.2000.”

5 CHOMSKY AND HALLE’S REVOLUTION IN PHONOLOGY

I would like to thank Morris Halle for discussing some of these issues with me,and Jim McGilvray for his detailed comments on an earlier version of this chapter.All errors of fact or interpretation are mine. I am grateful for the support of grants410–99–1309 and 410–2003–0913 from the Social Sciences and HumanitiesResearch Council of Canada.

1. For a detailed historical account of the context of twentieth-century phonologicaltheory see Anderson (1985).

2. A rule of the form A → B/C D is to be read, “A becomes B in the contextC D.” Parentheses indicate optional material. Thus, rule (1) indicates that a vowelbecomes [+long] before a glide followed by a voiced consonant or directly before avoiced consonant.

3. More precisely, Flapping may not precede the other two rules. In this case, the correctresults would obtain if all three rules applied simultaneously to the underlying form.There are many cases, however, in which simultaneous application does not succeed.

4. I abstract away from details and refinements such as the cycle.5. After a class of sounds called sibilants the plural has a third pronunciation, [əz], as

in busses, churches, bushes, and lounges.6. The relatively small number of features is a legacy of the work of Jakobson and

Halle in the 1950s, who sought to arrive at the smallest possible number of features



required to distinguish contrastive sounds. For a more recent survey of the phoneticparameters that can be used distinctively in the languages of the world see Ladefogedand Maddieson (1996). Though the range of variation seen across the spectrum ofthe world’s languages may be greater than was known in the 1950s, there still appearto be significant restrictions on how many of the possible contrasts may actually beemployed in a single language. See Rice (2002) for a discussion of vowel systems.

7. In one of the earliest works in generative grammar, Chomsky, Halle, and Lukoff(1956) show that an elegant analysis of English stress can be achieved by rules thatare sensitive to aspects of the syntax, in sharp contrast to previous treatments in theAmerican structuralist tradition. This work constitutes an early argument against thenotion that phonological analyses must be constructed without reference to otheraspects of grammar.

8. Hockett (1951) argues that the generalization concerning the distribution of voicelessand voiced consonants can be regained at the morphophonemic level, where wordboundaries and other morphological relations come into play. Thus, in his systemthe underlying morphophonemic representation of (9b) would be / / #patat#ak# / /.A morphophonemic rule of word-medial voicing would derive the (taxonomic)phonemic form / padatak /. The effect of this reshuffling of levels is to turn thetaxonomic phonemic level into a surrogate phonetic level, with the morphophonemiclevel doing much of the work formerly (and subsequently) assigned to the phonemiclevel. Once a systematic phonetic level is added to the model of grammar, there isno need for another level (the taxonomic phoneme) to play a similar role.

9. Of course, there could have been: we might have found, for example, that sip derivesfrom /ssip/ and zip derives from /sip/, and that the phonemic difference is between/s/ and /ss/, not /s/ and /z/. The analysis of a phonetic string is thus not possible inisolation, but must take into account other parts of the grammar.

10. Another well-known example is Swadesh and Voegelin’s analysis of the morpho-phonemics of Tubatulabal (Swadesh & Voegelin 1939). This work, done under theinfluence of Edward Sapir (1884–1939), is a testament to a more liberal period,before the neo-Bloomfieldian framework had considerably narrowed the range ofwhat was an acceptable analysis.

11. This interpretation of the status of morphophonemics was evidently not acceptedby Swadesh and Voegelin (1939): “If it has been possible . . . to reduce the apparentirregularity of Tubatulabal phonology to system, this very fact guarantees the truthof our theory.”

The question of what sort of reality was to be attributed even to phonemic analysisdid not receive a clear answer in American linguistic thought. Sapir argued unam-biguously that phonological analyses were “psychologically real” (Sapir 1933), butthis “mentalistic” interpretation was rejected by most American linguists (cf. Twad-dell 1935). Psychological realism was replaced by a recurring debate in Americanstructuralist linguistics as to whether an analysis should be thought of as “God’struth” or as “hocus-pocus” (see Joos 1957: 80). Chomsky (1957b) found thesediscussions to be “quite empty and sterile.”

12. See the contribution by Lasnik in this volume for further discussion of this argumentand the notion of levels in grammar. The notion has nevertheless persisted withingenerative phonology that there may be some fundamental difference between rulesthat deal only in contrastive feature values and rules that deal in redundant values.



One expression of this is the theory of Lexical Phonology and Morphology (Kiparsky1982a, 1985), which divides the phonology into lexical phonology (roughly, the oldmorphophonemic rules) and the postlexical phonology (mainly allophonic rules),The dividing line between lexical and postlexical phonology thus occupies a placesomewhat like that of the taxonomic phoneme, without, however, having to observethe old constraints on this level. A further difference is that proposals for a lexical–postlexical distinction in the phonological component are supported by empiricalevidence, by showing, for example, that the components have different properties.

13. Discovery procedures are not to be confused with the generative quest for explana-tory adequacy. The latter aims to account for how first language learners arrive atthe grammar of their language, given that the available evidence appears to greatlyunderdetermine the choice of grammar. Chomsky has proposed that learners areendowed with a rich theory of Universal Grammar that guides and limits theiracquisition of grammar. A theory of Universal Grammar differs from discoveryprocedures in that the latter are primarily intended to guide the linguist. Generativegrammar posits no procedures to guide the linguist in the construction of a theoryof Universal Grammar, nor are any conditions placed on how a theory of UniversalGrammar can be related to the facts of language. Of course, an important criterionin choosing between different linguistic theories is how well they fare in explaininghow language can be acquired.

14. This form of behaviorism is thus quite different from the classical empiricism ofHume and Locke. Though he believed in the priority of sense impressions over innateideas in the sense of Descartes, Hume saw his task as discovering the principles ofthe mind, including innate properties such as the imagination.

15. Chomsky (1959) demonstrated this in his famous review of B. F. Skinner’s Ver-bal Behavior (Skinner 1957). Chomsky showed that concepts that had concretemeanings in Skinner’s (1938) Behavior of Organisms, where they were appliedto classical but limited problems involving the relationship between stimuli andresponses in rats and pigeons, became either trivial or false when applied to humanverbal behavior.

16. Morris Halle continued, of course, to be at or near the center of developments inphonological theory for many years. In this essay, however, I focus only on his earlywork with Chomsky. See Halle (2002) for a collection of representative papers from1954 to 2002.

17. For recent introductions to contemporary phonology, see Gussenhoven and Jacobs(1998) and Roca and Johnson (1999). For more detailed treatments of phonologicaltheory up to the mid-1990s, see Kenstowicz (1994) and Goldsmith (1995).

18. The debate on abstractness was initiated by Kiparsky (1968). For a sampling of theextensive literature on the abstractness controversy from various points of view, seeDresher (1981a), Gussmann (1980), and Kenstowicz and Kisseberth (1977, ch. 1)on one side, and Hooper (1976), Linell (1979), and Tranel (1981) on the other.

19. In assessing learnability, it is relevant to ask what the learner already knows aboutthe grammar. Kaye (1974) comments that derivations that are opaque in Kiparsky’ssense may nevertheless aid a listener (and a learner) in recovering underlying forms,and hence the lexical identity of a morpheme. For example, the short raised diph-thong in [r�jɾər] (writer) in the dialects under discussion makes the Raising ruleopaque, but signals to a listener who already knows the rule that the flap derives



from a /t/. In a dialect where the rules apply in a different order so that writer soundsthe same as rider, the rules are less opaque, but the identity of the lexical items ismore difficult to recover.

20. For further discussion of this case see Dresher (1996).21. The SPE markedness theory was developed further by Kean (1980), but was oth-

erwise not much pursued in the years immediately after SPE. Some version ofmarkedness is found in most current approaches to phonology, albeit in differentforms. Calabrese (1995) presents a version that is much in the spirit of Kean andSPE. The theory of Government Phonology (Kaye, Lowenstamm & Vergnaud 1985)builds markedness into phonological representations. The same is true of ModifiedContrastive Specification (Avery & Rice 1989; Dresher, Piggott & Rice 1994).

6 UNIVERSAL ASPECTS OF WORD LEARNING

1. “Worth its salt” loosely translates as “an appropriate dissertation topic in linguistics.”In the end, of course, distributional analysis will distinguish even among these, e.g.“The one animal other than an elephant who trumpets is a ——” or “the largestextant land animal is the ———”.

2. This stipulation sounds innocent enough for the dog case, but is in fact not uncon-troversial. For example, in positions related to that first popularized by BenjaminWhorf, the causal chain for word learning is argued to run at least partly in theother direction, the learning of words being the generator or at least molder ofthe concepts. For discussion pro and con see Gentner and Goldin-Meadow (2003),Papafragou, Massey, and Gleitman (2003), inter alia.

3. The availability of grounding information is particularly important if “bootstrap-ping” approaches are not to be circular in the sense whimsically suggested by thevery term (one can’t lift oneself by one’s bootstraps if one is standing in the boots).Thus we must demonstrate that initial learning of the concrete nominal vocabu-lary need not invoke any syntactic or word-distributional knowledge (Gleitman &Gleitman 1997).

4. We should point out that the systematic relationship between core participant rolesin a conceptual/semantic structure and noun phrases in a sentence is not easy to seein every sentence. Mature speakers of most languages systematically leave out nounphrases when their referents are recoverable in the discourse context (e.g. Hangingtoo low). Immature speakers leave out noun phrases even when their referents cannotbe recovered, requiring us to look systematically at a large sample of their sentencesto uncover what they know about their own verbs (Bloom 1970). Likewise, thesentences of young home signers tend to be short, about two signs long. Acrossutterances, however, a particular child can be observed to sign all of the participantroles required by the logic of each verb – just not all at once (Feldman et al. 1978;Goldin-Meadow 2003).

7 EMPIRICISM AND RATIONALISM AS RESEARCHSTRATEGIES

I would like to thank Harry Bracken and Paul Pietroski for helpful comments on ear-lier drafts. I would especially like to thank Jim McGilvray for the detailed commentswithout which this chapter could not have been written.



1. Chomsky has been very active in identifying his precursors. These include manyless well known thinkers including the Cambridge Platonists (e.g. Cudworth) andvon Humboldt among others. See Chomsky (1966, 2002a) and McGilvray (1999) fordiscussion.

2. Knowledge is often analyzed as being justified true belief. In what follows I abstractaway from the issue of justification and concentrate on how true beliefs can arise.

3. The text should not be taken to suggest that any of this is unproblematic. Howabstraction is intended to function, how metrics of similarity are developed andwhere they come from is one of the problems for an empiricist theory of mind. Webriefly return to this issue below in our discussion of Chomsky’s views.

4. Similarity is a central empiricist notion. It was continually subjected to critical anal-ysis and revision. This is precisely as it should be as this is what allows empiricistepistemology to get off the ground. Of course, to the degree that the similarity notionfails, empiricism will as well.

5. For a more elaborate discussion of what is only lightly glossed here, see Lightfoot(this volume).

6. For a discussion, of this point, see Hornstein and Lightfoot (1981). For a more recentdiscussion, see Crain and Pietroski (2001).

7. Modern empiricist learning theories have attacked this problem, but it is questionablewhether they have entirely succeeded. The most recent sophisticated empiricist learn-ing theories are cast in a connectionist mold. See Marcus et al. (1999) and commen-tary by Pinker (1999) for arguments claiming that current connectionist architecturescannot adequately generalize to rules. If this is correct, then the problem noted under(b) is far from trivial for an empiricist learning theory.

8. The modern reader will likely replace God with Darwin and reach for evolutionaryaccounts to fill in the gap. There is no current reason for thinking that Darwinianapproaches will get one much further than Descartes’s, at least for interesting domainsof knowledge such as linguistic or mathematical knowledge.

8 INNATE IDEAS

1. This is not the trivial claim that “humans can acquire the languages they can acquire,”if only because of the substantive constraints on human languages that linguistshave discovered. Empirical inquiry suggests that the human faculty for language hasa distinctive character that is responsible for many features of human languages.But the mental state of “knowing Japanese” is not innate, any more than having awell-exercised heart is innate. What is innate, according to Chomsky, is the mentalsystem that makes it possible for any learner to acquire any human language – thelanguage faculty. Occasionally, innate resources may be damaged, through traumaor (independently specifiable) pathology. In this regard, the language faculty is nodifferent than the liver.

2. The process of going through puberty is governed largely by innate mechanisms;though even here, as Chomsky notes, environmental conditions matter. Adolescentswho are underfed won’t go through puberty, at least not in the normal way.

3. See Lasnik and McGilvray, both in this volume; see also Fodor (1998) and his reeval-uation of Fodor (1975).

4. Empiricism is, nonetheless, alive and well. See Pullum (1996), Bates and Elman(1996), Elman et al. (1996), Cowie (1999). For a recent Chomskyan reply, see Crain



and Pietroski (2001). It is often suggested that a proper understanding of inductionwill reveal language acquisition as a special case of induction. But not only is thisa mere promissory note, until someone explains how induction works, Goodman(1983) provides a compelling argument that induction is itself a form of humanknowledge acquisition to which poverty-of-simulus considerations apply. Oddly,this is not the conclusion Goodman drew from his argument; though it is certainlythe conclusion that Chomsky drew. And Chomsky has argued, on empirical grounds,that the underdetermination of linguistic knowledge by experience is not a specialcase of the underdetermination of inductive knowledge by experience. See Chomsky(1969a, 1986); cf. Quine (1960).

5. We suspect that this is why Chomsky tends to emphasize his connection withDescartes, rather than Kant. For while much of Chomsky’s nativism is surely inkeeping with Kantian themes, Kant himself tends to view (i) and (ii) as cons-trained – at least at some levels of description – by claims about how thinkers mustbe related to the world they inhabit in order to have any knowledge at all.

6. Moreover, showing someone a cow does not guarantee that he’ll see it as a cow. Anda rationalist might hold that while experience with cows can trigger (what comesto be) a person’s cow-concept, such experience does not (in any interesting sense)shape the resulting concept. But the rationalist’s claim is especially plausible withregard to ideas of things we can’t experience.

7. If a very limited experiential base suffices for “averaging” and discerning similarities“across” cases, the core empiricist notions cease to have any interesting meaning.Cf. Chomsky’s (1959) criticisms of extending behaviorist notions like “stimulus”and “response” from paradigms of operant conditioning to descriptions of verbalbehavior.

8. While it is unlikely that many subjects would be as quick as Meno, Plato illus-trates how quickly one can come to know a very abstract proposition; and thissuggests that such knowledge can be acquired with very little experience, giventhe right alternative “prompting” from a clever questioner. And if repeated expe-rience is not the only possible trigger for the knowledge, this reinforces the ratio-nalist claim: typical cases of knowledge acquisition, which involve a little moreexperience but less Socratic dialog, involve leaps beyond anything the sensesprovide.

9. A figure with area 2 is just a special case, though an especially interesting one.For had the dialog continued a bit, Meno could have been led to the idea of anirrational number (by thinking about the sides of a square with area 2). And what,in experience, provides the basis for that idea? At the risk of belaboring ancientmathematics, we also note that knowledge of infinite domains is presumably notacquired by experience; yet following Euclid, one can quickly show that there areinfinitely many primes. (If some number N were the largest prime, then N wouldbe the largest prime factor of N! + 1; but since N! + 1 is not evenly divisible byany integer less than or equal to N, N isn’t a factor of N! + 1). The broader point isthat just as an adequate human epistemology must do justice to our knowledge ofmathematics, so it must do justice to our knowledge of language.

10. But it would be a mistake to deny Plato’s conclusion on the grounds that he didn’tprovide a plausible account of how we come to have the cognitive resources thatmake mathematical knowledge possible. Similarly, it would have been a mistake todeny that humans have livers until someone explained how we come to have livers.



One can have excellent evidence that a species has certain traits independently ofexperience without yet knowing how they come to have those traits.

11. Whether such propositions are necessarily true is a complex matter, given non-Euclidean geometries (and the character of the physical universe). But in any case,even before seeing the proof, one knows that the Pythagorean Proposition is some-how necessarily correct if correct at all. And this fact about human thinkers is inde-pendent of any particular claim about what it is for a proposition to be necessarilycorrect.

12. As a candle burns it seems to vanish. We know, by investigation, that the matterdoes not cease to exist; it simply changes form. But its form changes enough thatour unaided senses can no longer track the matter we perceived as a candle.

13. Hume – arguably, the greatest empiricist to date – wrestled with this question (and thecorresponding questions about our ideas of necessity). And at least on one readingof Hume, endorsed by Chomsky, he grants the rationalist’s psychological point: wedo indeed have ideas that take us beyond experience; but for just that reason, weshould be skeptics (in the ancient sense of neither affirming nor denying) with regardto many judgments involving those ideas. But again, the question here is not whatcould justify judgments involving Cartesian ideas. We think Chomsky’s (2000a)views on semantics – and in particular, the irrelevance and futility of associatinglinguistic expressions with mind-independent referents – are usefully viewed in thislight; see McGilvray (1998, 1999, 2002a), Pietroski (2002, 2003) for discussion.

14. He has also speculated on the issue of moral knowledge. See, for example, Chomsky(1988b); see Dwyer (1999) for an attempt to develop the neo-Cartesian speculationsin more detail. These examples preserve analogs of the felt necessity that attachesto mathematical propositions. But it is an open question whether all knowledge thatstems from “the hand of nature” is associated with felt necessity.

15. One can make the point without appealing to English words. Let R be a set-theoreticrelation such that s1 bears R to s2 if and only if the set of things not in s1 is a subsetof the set of things not in s2. Any determiner � that labeled R would be a counterex-ample to (12). For � cows are brown would be true if and only if the set of thingsthat aren’t cows is a subset of the set of things that aren’t brown; and � cows arebrown cows would be true if and only if the set of things that aren’t cows is a subsetof the set of things that aren’t brown cows. So the following claim would be false:

if [(� cows)(are brown cows)], then [(� cows)(are brown)].

For this would say that if [the set of things that aren’t cows is a subset of the setof things that aren’t brown cows], then [the set of things which aren’t cows is asubset of the set of things that aren’t brown]; i.e. if everything that isn’t a cow isn’t abrown cow, then everything that isn’t a cow isn’t brown. But this is wrong. Trivially,everything that isn’t a cow isn’t a brown cow; but it hardly follows that everythingthat isn’t a cow isn’t brown. There are many other non-conservative set-theoreticrelations. This raises the question of why we evidently can’t use determiners toname relations like R, which (from a theoretical perspective) seems simpler thanmany relations named by complex determiners like more than seventeen but fewerthan twenty-six.

16. As this discussion makes clear, arguments for innateness exhibit a multi-prongedstructure. Initially, linguists consider an array of phenomena and propose someanalysis (in terms of linguistic categories and rules) as a partial explanation of



the phenomena. Such analysis is followed by cross-linguistic research, in searchof potential universals – candidates for inclusion in Universal Grammar. Givena potential linguistic universal, call it U, one can ask whether children could haveinferred that their language respects U on the basis of data available to them. If everychild acquires a grammar that respects U, but it is very implausible that every childencountered (and noticed) expressions that would have let them infer that their localgrammar respects U, this strongly suggests that U is a reflection of (innate) UniversalGrammar. If three-year-olds manifest knowledge of the very linguistic principlesthat characterize adult grammars, that compresses the learning problem consider-ably. This same strategy of providing converging (nondemonstrative) arguments forinnateness is followed in other domains, using similar criteria: early emergence,throughout the species, of traits that seem to be (dramatically) underdetermined byexperience. See Crain and Pietroski (2001) for further discussion.

17. Disjunction also seems to be “exclusive,” at least with regard to pragmatic implica-ture, in cases like Some cow ate broccoli or asparagus. While the sentence mightbe literally true if the only cow that ate broccoli or asparagus ate both, this is notthe natural expectation. Chierchia (2000) argues a (scalar) implicature of “exclusiv-ity” is computed when disjunction appears in a non-downward entailing linguisticcontext (such as in the second argument of every, and in both arguments of some),but not in downward entailing linguistic environments (such as the first argumentof every, or in either argument of no). See, Chierchia et al. (2001), for a semanticaccount of such implicature.

9 MIND, LANGUAGE, AND THE LIMITS OF INQUIRY

1. The standard papers here are Putnam (1975) and Burge (1979), plus Kripke’s classic(1972).

2. Fodor himself recognizes this Fregean problem and has wrestled very interestinglywith it over almost two decades and in many works right from the time he first for-mulated his naturalism about reference. See the next note for a crucial philosophicalpoint about Fodor’s views. Chomsky uses the term “perspective” sometimes to sim-ply raise the Fregean problem just raised by this first argument for denotationalviews of meaning. But he also goes on then to give a systematic, internalist sprucingup of the idea of “perspectives” which finds its place at an interface where use oflanguage picks up from. This sanitized notion of perspectives, however, is thereforenot intended to determine reference as traditional Fregean notions of sense are sup-posed to. In fact reference becomes irrelevant to the sanitized notion of perspectivessince it drops out of scientific linguistic study altogether for Chomsky. Again, seethe next note.

3. The second argument and its conclusion would also undercut any effort to solve theproblem created by the first argument, by appealing not to conceptions of things, butto purely syntactically or computationally described items, i.e. it would undercutany effort to solve the problem by appealing to anything which, unlike conceptionsof things, would not amount to a kind of content. And it is because conceptionsof things undeniably amount to a kind of content, that Fodor, in his keenness tosay that there is no content but denotational content, has often in his efforts todeal with the Frege puzzles made just such an appeal to purely syntactically andcomputationally described items. See for instance his (1987, 1998). Those are itemswhich may account for a misinformed thinker’s rationality, but do so by means



not available to the thinker, not self-known to him. Chomsky finds a place for hisnotion of “perspectives” in what he describes broadly as “syntax,” but in doingso he, unlike Fodor, gives up on the idea that perspectives (senses) are ways offeeding into reference, or solving anything like the problem created by the firstargument. Since he has no desire to make a naturalistic study of language bear onnotions of reference or intentional content, he can afford to do this, i.e. he can affordto ignore the self-knowledge constraint imposed by the second argument, whichFodor, because of precisely those aspirations, cannot. If Fodor were to reject theself-knowledge constraint on which the second argument above depends, it’s notclear, as the second argument points out, that he could fulfil the aspirations of beingcontinuous with an ineliminable intentional psychology. It may be that because hehas seen this point that Fodor himself in his most recent work on concepts seems tohave taken to stressing not so much that there will be syntactical and computationalsolutions to Fregean problems and puzzles raised for denotational semantics and thenaturalist intentional psychology it is supposed to yield, but rather stressing somesort of Leibnizian metaphysical conspiracy which ensures that Frege puzzles do notarise too often in our world, so they do not really pose a threat to such a conceptionof semantics and intentional psychology.

4. Kripke explicitly claims that senses will not solve Frege puzzles for just this reason.See his (1976).

5. We use the word “internal” here in a way that contrasts with external items such asplanets and cities. It is not intended to rule out the publicness or the abstractnessof internal objects such as senses which Frege insists on and which is what makesthem objects of thought in just the sense that Chomsky is attacking. To make themabstract objects is to make them fall afoul of the insight Chomsky invokes Reid anddu Marsais for, the insight provided in the self-knowledge constraint that the secondargument imposes.

6. See notes 2 and 3 for how the notion of perspectives sheds its traditional Fregeanassociation of sense (as determining reference) to an internalistically sanitized broadlysyntactic notion that has nothing directly to do with reference.

7. The argument that follows is stated most fully in Chomsky (2000a: ch. 4).8. Except that it exaggerates the distinction between philosophy and science, by pro-

jecting the distinction between academic disciplines in universities after Kant intothe past, when there was not such a great distinction – but only the sort of naturalphilosophy which was, then, both philosophy and science.

10 MEANING AND CREATIVITY

I am grateful to Noam Chomsky, Paul Pietroski, and Robert Stainton for reading andcommenting on this chapter.

1. Some authors (e.g. Steven Pinker) have assumed that its usefulness must be theproduct of Darwinian natural selection for communication – “history” of anothersort. Chomsky does not agree; see this volume’s introduction and Hauser, Chomsky& Fitch (2002).

2. This explains why people who acquire a language late in life often have difficultysounding like a native speaker. There does not seem to be a parallel difficulty withconcepts, although the case is not closed: perhaps different languages exploit different“semantic fields,” allowing for different ways to “cut up” conceptual space. For some



data, see Stephen Ullmann’s work – following Trier’s – on a “field” view of concepts(Ullmann 1957: 152f). The notion of a semantic field is sometimes attributed toHumboldt. It appears in several forms, e.g. Foucault’s conception of a grille – forhim, attributable to social power relations. It is possible to imagine several ways inwhich the nativist could accommodate it – parameterization, among others – shouldthere be empirical evidence that that is needed.

3. Landau and Gleitman (1985) show that blind children acquire color and sight vocab-ulary as easily as sighted. Blind children do not, of course, experience visual reds.

4. They also require a different “theory of meaning” than Chomsky’s, which dealsonly with natural languages.

5. By “root” concept I have in mind one that would – in English – constitute the corein common to (say) trivial, trivially, triviality, untrivial, trivialize. Adult speak-ers of English can be counted on to have acquired over 50,000 root concepts.

An account of linguistic concept acquisition need not explain acquisition ofconcepts in other faculties, nor explain how what might be called “cooperative” con-cepts are produced. Our commonsense understanding of the world clearly dependson more than language. For instance, bucket draws shape and color features fromvision – facts that Marr emphasized when he pointed out that humans with damagedlanguage centers retain a robust capacity to visually recognize buckets, althoughthey cannot say what they are for. Language contributes (perhaps) container,artifact , and used to move masses of non-rigid materials – thefunction of buckets, but not the abstract “shapes” Marr and others hold visioncontributes.

6. The basic points of “aitiational semantics” were Moravcsik’s (1975).7. Linguists have for decades focused on the broad architectural aspects of the language

faculty – on the structures of phrases, for example. Recently distributed morphol-ogists (Halle & Marantz 1993, 1994; Marantz 1997; Harley & Noyer 1999) haveabsorbed lexical categories such as N and V into syntax.

8. His recent work (2000b, 2001a) makes the point even more obvious, although itchanges the conception of derivation adopted in 1995c.

9. This is an informal suggestion; it is neither minimal nor “optimal” in Chomsky’ssense (1995c: 235). It does, though, meet two obvious constraints: that it pair“sound” (phonological) features with “meaning” ones (those that appear at SEM),and that it indicate (crudely) what will appear at SEM and the phonetic interface(PHON).

10. It can help to think of specific SEMs as events – as occasions where the languagefaculty yields a “perspective” that is employed by other systems in the head to“configure” experience. To conceive of them as events, think of them and theircontribution to cognition “adverbially” – as ways to experience, conceive, imagine,suggest, state etc.

11. Both may be the usual case, Chomsky suggests (1995c: 236). If so, ±concretewould be a principle of UG for all nouns and only exceptions would need markingin a lexical specification.

12. Putting it this way makes it clear why I-languages are so-called: they are Individually,Internally, Intensionally (Chomsky 1986) and – we might as well now add – Innatelyand Intrinsically specified languages.

13. It is not clear how to think of cooperation. For this chapter, think of it as jointconstitution of concepts.



14. These are insights emphasized by A. W. Schlegel and particularly Humboldt(Chomsky 1996a,b, 2002a).

15. Chomsky notes (1966: n. 9) that it is possible that Descartes was familiar with thework of Juan Huarte, who in his Examen de Ingenios in 1575 mentioned three levelsof creativity. He tied the (distinctively) human variety to having something like agenerative system or systems in the mind that is (are) detached from circumstance.

11 MARKET VALUES AND LIBERTARIAN SOCIALIST VALUES

1. Schlesinger doubted Alsop’s predictions of victory, but added, “We all pray that MrAlsop will be right” (The Bitter Heritage, cited in Chomsky (1971a: 240). In otherwords, Schlesinger’s opposition to the Vietnam War was based on his estimation ofits prospects for success, not his judgment of its moral or legal character. Chomskyobserved that the views of Schlesinger and Alsop “bound a substantial range ofAmerican opinion,” and that therefore it was of great importance to recognize that“each presents what can fairly be described as an apologia for American imperi-alism” (letter, Listener, January 15, 1970, p. 88). Schlesinger responded to thesecharges (January 29), and Chomsky had the last word (19 February).

2. Chomsky, of course, recognizes that our moral judgments “are obviously heavilyconditioned by various doctrinal systems with social and historical roots, and byperceived choices and available interpretations that are socially and historicallyconditioned” (1988a: 468).

3. Chomsky’s (1988a: 111) summary of Leibniz’s view.4. Confusion between these quite separate commitments is a fundamental flaw of, for

example, Wilkins (1997).5. Adam Smith, from The Theory of Moral Sentiments. Cited in Viner (1991: 90f).6. Cited in Dwyer (1998:65).7. For a puzzled contemplation of Smith’s bitter critique, see Oakley (1994: 105).8. Policy Planning Study (PPS) 23 of February 1948, cited in Chomsky (1987b: 15f).9. Smith (1990: 204); Malthus cited in Chomsky (1994: 186).

10. Tape, “At the Rowe Center,” Rowe, MA, 15–16 April 1989, tape 4, side A.

12 THE INDIVIDUAL, THE STATE, AND THE CORPORATION

1. Chomsky notes (1995b: 354) that the anarchist Bakunin coined this phrase.2. The Miringoffs (1999: 40) created an index of sixteen measurable social health fac-

tors, such as teenage suicides, unemployment, and wealth inequality, demonstratingthat America’s “social health” generally tracked the Gross National Product untilthe early 1970s. Since then, the GNP has continued to grow while overall socialhealth has significantly deteriorated.

3. An example from the UK is the Welsh development agency’s providing fundingon the order of a hundred million pounds to Korea’s LG Semicon for the purposeof building a semiconductor plant in Cardiff. The plant was abandoned after beingbuilt and it is not clear what happened to the funds granted.

4. Chomsky explains: “It’s not the case, as the naıve might think, that indoctrinationis inconsistent with democracy. Rather, as this whole line of thinkers [ReinholdNeibuhr, Walter Lippman] observes, it’s the essence of democracy” (quoted inAchbar 1994: 43).



5. Aside from lax enforcement, tax havens permit the wealthy to preserve their dispro-portionate power (Strange 1998: 131–7).

6. Chomsky discovered the phrase “de facto world government” in an article by JamesMorgan in the Financial Times, Weekend FT, April 25/26, 1992.

13 NOAM CHOMSKY: THE STRUGGLE CONTINUES

1. Draft of lecture entitled, “Control of Our Lives,” delivered in Albuquerque, NewMexico, Feb. 26, 2000.

2. See Chomsky (1996d, 1999b) for discussion of the peace of the “victors”; Znet, Fall2000 for Chomsky’s analysis of the Al Aqsa Intifadah.

14 THE RESPONSIBILITY OF THE INTELLECTUAL

I thank Edgard Andre, Xavier Bekaert, Jim McGilvray, and specially Diana Johnstonefor helpful comments on an earlier draft of this chapter.

1. Business Week, April 17, 2000. Patrick Smith (former correspondent of the Interna-tional Herald Tribune in Asia).

2. In the words of the “respected Roosevelt-Kennedy liberal Edward Bernays in hisclassic manual for the Public Relations industry, of which he was one of thefounders and leading figures”. Quoted in Chomsky (1999e); also available as “MarketDemocracy in a Neoliberal Order: Doctrines and Reality” (Davie Lecture, Univer-sity of Cape Town, May 1997) in Z Magazine, Sept. 1997: http://www.zmag.org/chomsky/articles.cfm.

3. Many topics that I will not cover here are analyzed in Chomsky’s works from thisviewpoint, including Central America, Israel, the “war on drugs,” terrorism, SouthAfrica, the Cold War and the arms race, etc.

4. It is interesting to note that Chomsky expressed his views on such topics long beforethe word “postmodernism” was invented. In 1966, in one of his first political essays,he contrasted his view of truth with Martin Heidegger’s who wrote in 1933 that “truthis the revelation of that which makes a people certain, clear and strong in its action andknowledge”. He also contrasted Heidegger’s view with the “more forthright” viewof Arthur Schlesinger who, after having explained that he had lied about the Bay ofPigs invasion, complimented the New York Times for having suppressed informationon that invasion “in the national interest” (1987a: 60). For other discussions of post-modernism by Chomsky, see his debate with Michel Foucault (1974a), the papersin http://www.zmag.org/ScienceWars/sciencechomreply.htm, and the exchange inRaskin & Bernstein (1987: 104–56).

5. Special issue on Science/Rationality from Z Magazine, available at:http://www.zmag.org/ScienceWars/sciencechomreply.htm

6. This is the first of a series of comparisons to be made in this chapter, between twosimilar situations, whose goal is to awaken the reader to the strangeness of someproposition or of some attitudes that are often implicitly taken for granted within thedominant discourse; this rhetorical device is frequently used by Chomsky.

7. This radical shift in methodology is a significant sign of an ideological and apologeticattitude.



8. Some of which will be discussed below. Concerning the functioning of the media,see inter alia Chomsky & Herman (1988) and Chomsky (1989). For the actions ofthe West during the post Cold War period, see Chomsky (1999e).

9. The expression comes from the following statement of the Secretary of State ofPresident Truman, Dean Acheson, made at the beginning of the Cold War: “Likeapples in a barrel infected by one rotten one, the corruption of Greece would infectIran and all to the east. It would also carry infection to Africa through Asia Minor andEgypt, and to Europe through Italy and France, already threatened by the strongestdomestic Communist parties in Western Europe.” Quoted in Chomsky (1987a: 211).For a more detailed analysis of recent history developing the “rotten apple” theme,see (1987b).

10. See e.g. the comments by George Kennan, quoted in Milan Rai (this volume).11. Starting with Chomsky (1969c). The expression “new mandarins” was borrowed

by Chomsky from Ithiel Pool, a political scientist at MIT who used it to refer tohimself and his colleagues.

12. See also Chomsky (1987a: 243) for criticism of Hannah Arendt’s views on theVietnam War being due to “neither power nor profit,” but rather to irrational factors,such as “image making.”

13. A similar remark can be made about all the criticisms of the form: “the U.S. pol-icy with respect to, say, Iraq or Cuba is a failure because it does not lead to theestablishment of democracy, of a lasting peace, etc.” That kind of criticism iseasy and therefore appealing to many but, by taking at face value the proclaimedgoals of the rulers, it only strengthens the illusion that those goals are the realones. If one considers what are presumably the real goals, e.g. preventing therot to spread, controlling the oil reserves, etc., these policies are actually quitesuccessful.

14. See Chomsky (1996a,b: 64), in particular, for criticism of the position taken by theAmerican journal Dissent.

15. Chomsky (1987a: 144–5); on another occasion, Chomsky quoted Bertrand Russell’sremarks on the Russian Revolution: “every failure of industry, every tyrannous reg-ulation brought about by the desperate situation, is used by the Entente as a justifi-cation of its policy. If a man is deprived of food and drink, he will grow weak, losehis reason, and finally die. This is not usually considered a good reason for inflictingdeath by starvation. But where nations are concerned, the weakness and strugglesare regarded as morally culpable and are held to justify further punishment . . . Is itsurprising that professions of humanitarian feeling on the part of the English peopleare somewhat coldly received in Soviet Russia?” (Chomsky 1971b); the quote isfrom (Russell 1920). Similar remarks could be made about Iran or North Koreatoday.

16. See Chomsky (1987a: 326) for examples of remarkable cruelty: the United Statestried to prevent India from sending one hundred buffalo to Vietnam and AmericanMennonites from sending pencils to Cambodia.

17. See e.g. “Chomsky on Cuba,” answer to a question, posted on http://www.zmag.org/chomsky/index.cfm.

18. For a more extended discussion, see Chomsky (1988a: 204–8).19. For a detailed refutation of those and similar charges: Herman (1993).20. The following story illustrates the permanence of certain methods of psychological

pressure against pacifists and anti-interventionists. When in 1916 Bertrand Russell



managed to send an appeal to United States President Woodrow Wilson urging himto work in favor of a negotiated peace in Europe, he received from his long-timefriend and collaborator Alfred North Whitehead (who, unlike Russell, supported thewar) a newspaper report about German misdeeds against the French and Belgians,which Whitehead blamed on the “damping down among neutrals – America inparticular – of the first protests against earlier atrocities.” Adding, “what are yougoing to do to help these people”; see Monk (1997: 487).

21. For more details and references to original sources concerning this section, seeChomsky (1998). Available from: http://www.zmag.org/chomsky/articles.cfm.

22. This notion of subsidy is itself an interesting ideological construct, since it assumesthat the form of trade (including maybe the international debt and its trappingmechanisms) between the rich countries and the poor ones are somehow “just,”while the price stability and the guaranteed purchases offered by the Soviet Unionto Cuba were a form of “subsidy.”

23. From “Intellectuals and the State” (1977), reprinted in Chomsky (1982).24. Since here motivations such as fear of the draft do not play any role.


References

Achbar, Mark (1994). Manufacturing Consent: Noam Chomsky and the Media.Montreal: Black Rose.

Anderson, Stephen R. (1985). Phonology in the Twentieth Century: Theories of Rulesand Theories of Representations. Chicago: University of Chicago Press.

Anderson, S. R. & D. W. Lightfoot (2000). “The Human Language Faculty as an Organ.”Annual Review of Physiology 62.

Aristotle (1984). Nicomachean Ethics. Princeton: Princeton University Press.(1990). The Politics of Aristotle. Trans. Jowett. New York: Colonial Press.

Avery, Peter & Keren Rice (1989). “Segment Structure and Coronal Underspecification.”Phonology 6: 179–200.

Baker, M. (2001). The Atoms of Language. New York: Basic Books.Baker, S. A., W. J. Isardi, R. M. Golinkoff & L. A. Petitto (in press). “The Discovery of

Phonetic Forms.” Memory & Cognition.Baker, S. A., J. Sootsman, R. M. Golinkoff & L. A. Petitto (2003). “Hearing Four-

month-olds’ Perception of Handshapes in American Sign Language: No ExperienceRequired.” Published Abstracts for the Society for Research in Child Development2003 Biennial Meeting, Tampa, FL.

Barsky, Robert (1997). Noam Chomsky: A Life of Dissent. Cambridge, MA: MIT Press.Barwise, J. & R. Cooper (1981). “Generalized Quantifiers and Natural Language.”

Linguistics and Philosophy 4: 159–219.Bates, E. & J. Elman (1996). “Learning Rediscovered.” Science 274: 1849–50.Bellugi, U., S. Marks, A. Bihrle & H. Sabo (1993). “Dissociation Between Language

and Cognitive Functions in Williams Syndrome.” In D. Bishop & K. Mogford, eds.Language Development In Exceptional Circumstances. Hillsdale, NJ: LawrenceErlbaum Associates.

Bickerton, D. (1999). “How to Acquire Language without Positive Evidence: WhatAcquisitionists Can Learn from Creoles.” In DeGraff (1999).

Bishop, D. V. M. (1997). Uncommon Understanding: Development and Disorders ofLanguage Comprehension in Children. London: Psychology Press.

Bloch, Bernard (1941). “Phonemic Overlapping.” American Speech 16: 278–84.Reprinted in Joos (1957), 93–6. Also in Makkai (1972), 66–70.

Bloom, L. (1970). Language Development: Form and Function in Emerging Grammars.Cambridge, MA: MIT Press.

Bloom, P. (2000). How Children Learn the Meanings of Words. Cambridge, MA: MITPress.

Borer, H. (1984). Parametric Syntax. Dordrecht: Foris.

313


314 References

Bowerman, M. (1987). “The ‘No Negative Evidence’ Problem: How Do Children AvoidConstructing an Overly General Grammar.” In J. Hawkins, ed. Explaining Lan-guage Universals. Oxford: Blackwell. 73–101.

Brent, M. R. (1994). “Surface Cues and Robust Inference as a Basis for the EarlyAcquisition of Subcategorization Frames.” In L. R. Gleitman & B. Landau, eds.The Acquisition of the Lexicon. Cambridge, MA: MIT Press. 433–70.

Bresnan, Joan W. (1971). “Sentence Stress and Syntactic Transformations.” Language47: 257–81.

Burchill, Scott & Noam Chomsky (1998). “Human Nature, Freedom, and PoliticalCommunity: An Interview with Noam Chomsky.” Citizenship Studies 2 (1): 5–21.

Burge, Tyler (1979). “Individualism and the Mental.” In P. French, T. Uehling &H. Wettstein, eds. Midwest Studies in Philosophy 6. Minneapolis: University ofMinnesota Press.

Calabrese, Andrea (1995). “A Constraint-based Theory of Phonological Markednessand Simplification Procedures.” Linguistic Inquiry 26: 373–463.

Callanan, M. A. (1985). “How Parents Label Objects for Young Children: The Role ofInput in the Acquisition of Category Hierarchies.” Child Development 56: 508–23.

Carey, S. (1978). “The Child as Word Learner.” In M. Halle, J. Bresnan & G. A. Miller,eds. Linguistic Theory and Psychological Reality. Cambridge, MA: MIT Press.264–93.

Charron, F. & L. A. Petitto (1991). “Les premiers signes acquis par des enfants sourdsen Langue des Signes Quebecoise (LSQ): Comparaison avec les premiers mots.”Revue Quebecoise de Linguistique Theorique et Appliquee. 10 (1): 71–122.

Chierchia, Gennaro (2000). “Scalar Implicatures and Polarity Items.” Paper presentedat WELS 31, Georgetown University, Washington DC.

(2001). “Scalar Implicatures, Polarity Phenomena, and the Syntax/PragmaticsInterface.” Ms, University of Milan, Bicocca.

Chierchia, Gennaro, Stephen Crain, Maria Teresa Guasti & Rosalind Thornton (1998).“‘Some’ and ‘Or’: A Study on the Emergence of Logical Form.” In A. Greenhill,M. Hughes, H. Littlefield & H. Walsh, eds. Proceedings of the Boston UniversityConference on Language Development 11: 97–108. Sommerville, MA: CascadillaPress.

Chierchia, G., S. Crain, M. T. Guasti, A. Gualmini & L. Meroni (2001). “The Acqui-sition of Disjunction: Evidence for a Grammatical View of Scalar Implicatures.”Proceedings of the 25th Annual Boston University conference on child languagedevelopment. Somerville, MA: Cascadilla Press.

Chierchia, G. & S. McConnell-Ginet (2000). Meaning and Grammar, 2nd edition.Cambridge, MA: MIT Press.

Choi, S. & Bowerman, M. (1991). “Learning to Express Motion Events in English andKorean: The Influence of Language-specific Lexicalization Patterns.” Cognition41: 83–121.

Chomsky, N. (1951). Morphophonemics of Modern Hebrew. Masters thesis, Universityof Pennsylvania, Philadelphia. [Published 1979 New York: Garland Press.]

(1955). The Logical Structure of Linguistic Theory. Ms. Harvard University,Cambridge, MA and MIT, Cambridge, MA. [Published in part by New York:Plenum, 1975; Chicago: University of Chicago Press, 1985.]


References 315

(1957a). Syntactic Structures. The Hague: Mouton.(1957b). Review of R. Jakobson and M. Halle, Fundamentals of Language, Interna-

tional Journal of American Linguistics 23: 234–41.(1959). “A Review of B. F. Skinner, Verbal Behavior.” Language 35: 26–58. Reprinted

in Fodor & Katz (1964), 547–78, and, with added preface, in Jakobovitz & Miron,eds. Readings in the Psychology of Language. Englewood Cliffs, NJ: Prentice-Hall.142–72.

(1964). Current Issues in Linguistic Theory. The Hague: Mouton. Also in Fodor &Katz (1964), 50–118.

(1965). Aspects of the Theory of Syntax. Cambridge, MA: MIT Press.(1966). Cartesian Linguistics. New York: Harper and Row. [2nd edition: 2002a].(1969a). “Quine’s Empirical Assumptions.” In D. Davidson & J. Hintikka, eds. Words

and Objections: Essays on the Work of W. V. Quine. Dordrecht: Reidel.(1969b). At War With Asia. New York: Pantheon.(1969c). American Power and the New Mandarins. Harmondsworth: Penguin.(1970). “Deep Structure, Surface Structure, and Semantic Interpretation.” In Roman

Jakobson & Shigeo Kawamoto, eds. Studies in General and Oriental LinguisticsPresented to Shiro Hattori on the Occasion of his Sixtieth Birthday. Tokyo: TECCompany, Ltd. 52–91.

(1971a). American Power and the New Mandarins. London: Penguin/Pelican.(1971b). Problems of Knowledge and Freedom. The Russell Lectures. New York:

Pantheon.(1971c). “Foreword.” Prevent the Crime of Silence (reports from the sessions of the

International War Crimes Tribunal). Bertrand Russell Peace Foundation, Ltd. Avail-able at: http://www.911review.org/Wget/www.homeusers.prestel.co.uk/littleton/vltribun.htm

(1972a). Language and Mind. New York: Harcourt, Brace, Jovanovitch (expandedversion of 1968 edition).

(1972b). “The Pentagon Papers as Propaganda and as History.” In Noam Chomsky& Howard Zinn, eds. The Pentagon Papers: The Senator Gravel Edition, vol. V:Critical Essays. Boston: Beacon Press.

(1973a). “Conditions on Transformations.” In Stephen Anderson & Paul Kiparsky,eds. A Festschrift for Morris Halle. New York: Holt, Rinehart and Winston. 232–86.

(1973b). For Reasons of State. New York: Pantheon.(1974a). “Human Nature: Justice versus Power.” [A debate with Michel Foucault.] In

Fons Elders, ed. Reflexive Waters. Toronto: J. M. Dent.(1974b). Peace in the Middle East? New York: Pantheon.(1975). Reflections on Language. New York: Pantheon.(1977a). Essays on Form and Interpretation. New York: North-Holland.(1977b). Language and Responsibility: Conversations with Mitsou Ronat. New York:

Pantheon.(1977c) “On Wh-Movement.” In P. Culicover, T. Wasow & A. Akmajian, eds. Formal

Syntax. New York: Academic Press.(1980). Rules and Representations. Oxford: Blackwell.(1981a). Lectures on Government and Binding. Dordrecht: Foris.(1981b). “Principles and Parameters in Syntactic Theory.” In Hornstein & Lightfoot

(1981), 123–46.


316 References

(1981c). Radical Priorities, ed. Carlos Otero. Montreal: Black Rose.(1982). Towards a New Cold War: Essays on the Current Crisis and How We Got

There. New York: Pantheon.(1983). The Fateful Triangle: The United States, Israel & the Palestinians. Boston:

South End. [2nd edition: 1999b.](1984). Modular Approaches to the Study of Mind. San Diego: San Diego State

University Press.(1985). Turning the Tide. Boston: South End.(1986). Knowledge of Language: Its Nature, Origin, and Use. New York: Praeger.(1987a). The Chomsky Reader, ed. James Peck. New York: Pantheon.(1987b). On Power and Ideology: The Managua Lectures. Boston: South End.(1988a). Language and Politics, ed. Carlos Otero. Montreal: Black Rose.(1988b). Language and Problems of Knowledge. Cambridge, MA: MIT Press.(1989). Necessary Illusions: Thought Control in Democratic Societies. London: Pluto;

Toronto: Anansi.(1991). Deterring Democracy. London and New York: Verso.(1992a). “A View From Below.” Diplomatic History 16 (1).(1992b) Chronicles of Dissent: Interviews with David Barsamian. Stirling: AK Press.(1993a). “A Minimalist Program for Linguistic Theory.” In Kenneth Hale & Samuel J.

Keyser, eds. The View from Building 20: Essays in Linguistics in Honor of SylvainBromberger. Cambridge, MA: MIT Press.

(1993b). Language and Thought. Wakefield, RI: Moyer Bell.(1993c). The Prosperous Few and the Restless Many. Berkeley: Odonian.(1993d). Year 501: The Conquest Continues. Boston: South End.(1994). World Orders, Old and New. London: Pluto. [US edition: 1996d.](1995a). “Categories and Transformations.” In The Minimalist Program. Cambridge,

MA: MIT Press. 219–394.(1995b). “Language and Nature.” Mind 104: 1–61.(1995c). The Minimalist Program. Cambridge, MA: MIT Press.(1996a). Perspectives on Power. Montreal: Black Rose.(1996b). Powers and Prospects. Boston: South End. [US edition of 1996a.](1996c). “Anarchism, Marxism and Hope for the Future.” Red & Black Revolution,

no. 2. (Available at http://www.zmag.org/chomsky/index.cfm)(1996d). World Orders, Old and New. New York: Columbia University Press. [US

edition of 1994.](1997). “Serial Veto.” Index on Censorship 6.(1998). “The United States and the ‘Challenge of Relativity.’” In Tony Evans, ed.

Human Rights Fifty Years On: A Reappraisal. New York: St. Martins. (Availableat http://www.zmag.org/chomsky/articles.cfm)

(1999a). “Derivation by Phase.” In MIT Occasional Papers in Linguistics 18. Also inM. Kenstowicz, ed. Ken Hale: A Life in Language. Cambridge, MA: MIT Press,2001.

(1999b). The Fateful Triangle. Cambridge, MA: South End. [2nd edition of 1983.](1999c). “Language and the Brain”. Address at the European Conference on Cognitive

Science, October 1999, Siena. [Reprinted in 2002b.](1999d). The New Military Humanism. Monroe, ME: Common Courage.


References 317

(1999e). Profit Over People: Neoliberalism and Global Order. New York: SevenStories.

(1999f). “The Umbrella of U.S. Power: The Universal Declaration of Human Rightsand the Contradictions of U.S. Policy.” Ms.

(2000a). New Horizons in the Study of Language and Mind, ed. Neil Smith.Cambridge: Cambridge University Press.

(2000b). “Minimalist Inquiries: The Framework.” In Roger Martin, David Michaels& Juan Uriagereka, eds. Step by Step: Essays on Minimalist Syntax in Honor ofHoward Lasnik. Cambridge, MA: MIT Press. 89–155.

(2000c). “US Iraq Policy: Motives and Consequences.” In A. Arnove, ed. Iraq UnderSiege. Cambridge, MA: South End.

(2000d). A New Generation Draws the Line: Kosovo, East Timor and the Standardsof the West. London and New York: Verso.

(2000e). Rogue States. Boston: South End.(2001a). “Beyond Explanatory Adequacy.” Ms., Department of Linguistics, MIT. To

appear in A. Belletti, ed., Structures and Beyond: Current Issues in the Theory ofLanguage (in prep.).

(2001b). 9–11. New York: Seven Stories.(2002a). Cartesian Linguistics, ed. with a new introduction by James McGilvray.

Christchurch, NZ: Cybereditions. [2nd edition of 1966; includes new translations,index, and bibliography. Available at http://www.cybereditions.com]

(2002b). On Nature and Language. Cambridge: Cambridge University Press.(2003). Hegemony or Survival. New York: Metropolitan Books.

Chomsky, N. & David Barsamian (1992). Chronicles of Dissent: Interviews With DavidBarsamian. Monroe, ME: Common Courage.

(1994). Keeping the Rabble in Line: Interviews With David Barsamian. Monroe, ME:Common Courage.

(1998). The Common Good. Monroe, ME: Common Courage.(1996). Class Warfare: Interviews With David Barsamian. Monroe, ME: Common

Courage.Chomsky, N. & Morris Halle (1968). The Sound Pattern of English. New York: Harper

& Row.Chomsky, N., Morris Halle & Fred Lukoff (1956). “On Accent and Juncture in English.”

In Morris Halle et al., eds. For Roman Jakobson: Essays on the Occasion of hisSixtieth Birthday. The Hague: Mouton. 65–80.

Chomsky, N. & Suzy Hansen (2001). Interview in Salon.com: “Noam Chomsky:The Nation’s Most Implacable Critic of U.S. Foreign Policy Argues that theWar is Unjust, America is the Biggest Terrorist State and Intellectuals AlwaysSupport Official Violence.” Jan 16. (http://monkeyfist.com/ChomskyArchive/interviews/salon.html).

Chomsky, N. & Edward S. Herman (1979a). The Washington Connection and ThirdWorld Fascism, vol. 1 of The Political Economy of Human Rights. Boston: SouthEnd.

(1979b). After the Cataclysm: Postwar Indochina and the Reconstruction of Imperi-alist Ideology, vol. 2 of The Political Economy of Human Rights. Boston: SouthEnd.


318 References

(1988). Manufacturing Consent: The Political Economy of the Mass Media. NewYork: Pantheon.

Clahsen, H., S. Bartke & S. Gollner (1997). “Formal Features in Impaired Grammars:A Comparison of English and German SLI Children.” Essex Research Reports inLinguistics 14: 42–75.

Clark, E. V. (1990). “Speaker Perspective in Language Acquisition.” Linguistics 28:1201–20.

Cole, P. & G. Hermon (1981). “Subjecthood and Islandhood: Evidence from Quechua.”Linguistic Inquiry 12:1–30.

Cormack, A. & N. V. Smith (2000). “Fronting: The Syntax and Pragmatics of ‘Focus’and ‘Topic.’” UCL Working Papers in Linguistics 12.

Cowie, F. (1999). What’s Within: Nativism Reconsidered. New York: Oxford.Crain, S. (1991). “Language Acquisition in the Absence of Experience.” Behavioral and

Brain Sciences 14: 597–612.Crain, Stephen & Paul Pietroski (2001). “Nature, Nurture, and Universal Grammar.”

Linguistics and Philosophy 24: 139–86.Crain, S. & R. Thornton (1998). Investigations in Universal Grammar: A Guide to

Experiments on the Acquisition of Syntax and Semantics. Cambridge, MA: MITPress.

Croft, W. (1990). Typology and Universals. Cambridge: Cambridge UniversityPress.

Cudworth, Ralph (1995 [1737 (1688)]). A Treatise Concerning Eternal and ImmutableMorality, ed. Hutton. Cambridge: Cambridge University Press.

Curran, J. & J. Seaton (1985). Power Without Responsibility: The Press and Broadcastingin Britain, 2nd edition. London: Methuen.

Davidson, Donald (1967). “Truth and Meaning.” Synthese 17.Dawkins, R. (1976). The Selfish Gene. Oxford: Oxford University Press.DeGraff, M. (ed.) (1999). Language Creation and Change: Creolization, Diachrony and

Development. Cambridge, MA: MIT Press.Descartes, Rene (1984). Discourse. In Cottingham, Stoothoff & Murdoch, trans. The

Philosophical Writings of Descartes. Cambridge: Cambridge University Press.Dobzhansky, T. (1970). Genetics of the Evolutionary Process. New York: Columbia

University Press.Dowty, D. (1991). “Thematic Proto-roles and Argument Selection.” Language 67: 547–

619.Dresher, B. Elan (1981a). “Abstractness and Explanation in Phonology.” In Hornstein

& Lightfoot (1981), 76–115.(1981b). “On the Learnability of Abstract Phonology.” In C. L. Baker & John J.

McCarthy, eds. The Logical Problem of Language Acquisition. Cambridge, MA:MIT Press. 188–210.

(1985). Old English and the Theory of Phonology. New York: Garland.(1996). “Learnability and Phonological Theory.” In Jacques Durand & Bernard Laks,

eds. Current Trends in Phonology: Models and Methods 1: 245–66. Manchester:European Studies Research Institute; University of Salford Publications.

Dresher, B. Elan, Glyne L. Piggott & Keren Rice (1994). “Contrast in Phonology:Overview.” Toronto Working Papers in Linguistics 13 (1): iii–xvii.


References 319

Dretske, Fred (1981). Knowledge and the Flow of Information. Cambridge, MA: MITPress.

Dwyer, John (1998). An Age of the Passions: An Interpretation of Adam Smith andScottish Enlightenment Culture. East Linton: Tuckwell Press.

Dwyer, Susan (1999). “Moral Competence.” In K. Murasagi & R. Stainton, eds. Phi-losophy and Linguistics. Boulder, CO: Westview.

Edwards, S. & R. Bastiaanse (1998). “Diversity in the Lexical and Syntactic Abilitiesof Fluent Aphasic Speakers.” Aphasiology 12 (2): 99–117.

Elman, J. L., E. Bates, M. H. Johnson, A. Karmiloff-Smith, D. Parisi & K. Plunkett(1996). Rethinking Innateness: A Connectionist Perspective on Development.Cambridge, MA: MIT Press.

Epstein, Samuel D. (1999). “Un-principled Syntax: The Derivation of SyntacticRelations.” In Samuel D. Epstein & Norbert Hornstein, eds. Working Minimalism.Cambridge, MA: MIT Press. 317–45.

Feldman, H., S. Goldin-Meadow & L. R. Gleitman (1978). “Beyond Herodotus: TheCreation of Language by Linguistically Deprived Deaf Children.” In A. Lock, ed.Action, Symbol, and Gesture: The Emergence of Language. New York: AcademicPress. 351–414.

Fernald, A. (2003). “How 2-year-olds Look as They Listen: The Search for the ObjectBegins at the Verb.” Paper presented at the Biennial meeting of the Society forResearch on Child Development, Tampa, FL.

Fillmore, C. J. (1968). “The Case for Case.” In E. Bach & R. T. Harms, eds. Universalsin Linguistic Theory. New York: Holt, Rinehart and Winston. 1–88.

Fisher, C. (1996). “Structural Limits on Verb Mapping: The Role of Analogy inChildren’s Interpretations of Sentences.” Cognitive Psychology 31: 41–81.

(2000a). “From Form to Meaning: A Role for Structural Analogy in the Acquisitionof Language.” In H. W. Reese, ed. Advances in Child Development and Behavior27: 1–53. New York: Academic.

(2000b). “Who’s Blicking Whom? Word Order Influences Toddlers’ Interpretationsof Novel Verbs.” Paper presented at the Biennial International Conference on InfantStudies, Brighton, England.

(2002). “Structural Limits on Verb Mapping: The Role of Abstract Structurein 2.5-year-olds’ Interpretations of Novel Verbs.” Developmental Science 5:56–65.

Fisher, C. & L. R. Gleitman (2002). “Language Acquisition.” In H. F. Pashler (series ed.)and C. R. Gallistel (volume ed.) Stevens’ Handbook of Experimental Psychology,vol. 3: Learning and Motivation. New York: Wiley. 445–96.

Fisher, C., L. R. Gleitman & H. Gleitman (1991). “On the Semantic Content of Subcat-egorization Frames.” Cognitive Psychology 23: 331–92.

Fisher, C. & J. Snedeker (2002). “Counting the Nouns: Simple Sentence-structure CuesGuide Verb Learning in 21-month-olds.” Paper presented at the Boston UniversityConference on Language Development, Boston, MA.

Fodor, J. A. (1975). Psychosemantics. Cambridge, MA: MIT Press.(1983). The Modularity of Mind. Cambridge, MA: MIT Press.(1987). Psychosemantics: The Problem of Meaning in the Philosophy of Mind.

Cambridge, MA: MIT Press.


320 References

(1990). “Substitution Arguments and the Individuation of Belief.” In Fodor, A Theoryof Content and Other Essays. Cambridge, MA: MIT Press.

(1998). Concepts. Cambridge, MA: MIT Press.(2000). The Mind Doesn’t Work That Way: Scope and Limits of Computational

Psychology. Cambridge, MA: MIT Press.Fodor, J. A. & J. J. Katz (eds.) (1964). The Structure of Language. Englewood Cliffs,

NJ: Prentice-Hall.Frege, Gottlob (1892). “Uber Sinn und Bedeutung.” Zeitschrift fur Philosophie und

Philosophische Kritik 100.Freundlich, Fred (2000). The Mondragon Cooperative Corporation (MCC): An

Introduction. Ownership Associates Resource Library (http://www.ownershipassociates.com).

Friedmann, N. & Y. Grodzinsky (1997). “Tense and Agreement Agrammatic Production:Pruning the Syntactic Tree.” Brain and Language 56: 397–425.

Frith, U. (1991). Autism and Asperger Syndrome. Cambridge: Cambridge UniversityPress.

Fromkin, V. (1973). Speech Errors as Linguistic Evidence. The Hague: Mouton.(1996). “Some Thoughts About the Brain/Mind/Language Interface.” Lingua 100:

3–27.Froud, K. (2000). “Prepositions and the Lexical/Functional Divide: Aphasic Evidence.”

Lingua.Garnsey, S. M., N. J. Pearlmutter, E. Myers & M. A. Lotocky (1997). “The Contributions

of Verb Bias and Plausibility to the Comprehension of Temporarily AmbiguousSentences.” Journal of Memory & Language 37: 58–93.

Gentner, D. (1982). “Why Nouns Are Learned Before Verbs: Linguistic Relativity VersusNatural Partitioning.” In K. Bean, ed. Language, Thought, & Culture. Hillsdale,NJ: Erlbaum. 301–34.

Gentner, D. & L. Boroditsky (2001). “Individuation, Relativity and Early Word Learn-ing.” In M. Bowerman & S. Levinson, eds. Language Acquisition and ConceptualDevelopment. New York: Cambridge University Press. 215–56.

Gentner, D. & S. Goldin-Meadow (2003). Language in Mind. Cambridge, MA: MITPress.

Gillette, J., H. Gleitman, L. R. Gleitman & A. Lederer (1999). “Human Simulations ofVocabulary Learning.” Cognition 73: 135–76.

Gleitman, L. R. (1990). “The Structural Sources of Verb Meanings.” Language Acqui-sition 1: 3–55.

Gleitman, L. & H. Gleitman (1997). “What Is a Language Made Out Of?” Lingua 100:29–55.

Gleitman, L. R., H. Gleitman, B. Landau, & E. Wanner (1988). “Where Learning Begins:Initial Representations for Language Learning.” In F. J. Newmeyer, ed. Linguistics:The Cambridge Survey, vol. III: Language: Psychological and Biological Aspects.New York: Cambridge University Press. 150–93.

Gleitman, L. R., H. Gleitman, C. Miller & R. Ostrin (1996). “Similar, and SimilarConcepts.” Cognition 58: 321–76.

Goldberg, A. E. (1995). Constructions: A Construction Grammar Approach to ArgumentStructure. Chicago: Chicago University Press.

Goldin-Meadow, S. (2003). The Resilience of Language. New York: Psychology Press.


References 321

Goldsmith, John A. (ed.) (1995). The Handbook of Phonological Theory. Oxford: Black-well.

Goodman, J. C., L. McDonough & N. B. Brown (1998). “The Role of Semantic Contextand Memory in the Acquisition of Novel Nouns.” Child Development 69: 1330–44.

Goodman, Nelson (1983). Fact, Fiction, and Forecast, 4th edition. Cambridge, MA:Harvard University Press.

Gopnik, M. (1990) “Feature Blindness: A Case Study.” Language Acquisition 1 (2):139–64.

(1997) “Language Deficits and Genetic Factors.” Trends in Cognitive Sciences 1: 5–9.Gopnik, M. & M. Crago (1991). “Familial Aggregation of a Developmental Language

Disorder.” Cognition 39: 1–50.Gordon, P. & J. Chafetz (1990). “Verb-based versus Class-based Accounts of Actionality

Effects in Children’s Comprehension of Passives.” Cognition 36: 227–54.Gould, S. J. & R. Lewontin (1979). Proceedings of the Royal Society of London 205:

581–98.Grimshaw, J. (1990). Argument Structure. Cambridge, MA: MIT Press.Gussenhoven, Carlos & Haike Jacobs (1998). Understanding Phonology. London:

Arnold.Gussmann, Edmund (1980). Studies in Abstract Phonology. Cambridge, MA: MIT Press.Haider, Ejaz (2001). “Why Are We Flocking to Hear Chomsky?” The Friday Times.

Nov. 25 (http://www.thefridaytimes.com).Haley, Michael C. and Ronald Lunsford (1994). Noam Chomsky. New York: Twayne

Publishers.Halle, Morris (1959). The Sound Pattern of Russian: A Linguistic and Acoustical Inves-

tigation. [Revised version of 1955 Harvard University Ph.D. dissertation.] TheHague: Mouton, 2nd printing, 1971.

(2002). From Memory to Speech and Back: Papers on Phonetics and Phonology1954–2002. Berlin: Mouton de Gruyter.

Halle, Morris & Alec Marantz (1993). “Distributed Morphology and the Pieces of Inflec-tion.” In K. Hale & S. J. Keyser, eds. The View from Building 20. Cambridge, MA:MIT Press.

(1994). “Some Key Features of Distributed Morphology.” MITWPL 21: Papers onPhonology and Morphology. Cambridge, MA: MITWPL.

Harley, Heidi & Rolf Noyer (1999). “Distributed Morphology.” Glot International 4 (4):3–9.

Harris, Zellig (1951). Methods in Structural Linguistics. Chicago: University of ChicagoPress.

Hauser, M. D., N. Chomsky, & W. T. Fitch (2002). “The Faculty of Language: What IsIt, Who Has It, and How Did It Evolve?” Science 298: 1569–79.

Herman, Edward S. (1993). “Pol Pot, Faurisson, and the Process of Derogation.” InNoam Chomsky: Critical Assessments, ed. Carlos Otero. London: Routledge.

Hirsh-Pasek, K. & R. M. Golinkoff (1996). The Origins of Grammar: Evidence FromEarly Language Comprehension. Cambridge, MA: MIT Press.

Hockett, Charles F. (1951). Review of Phonology as Functional Phonetics by A.Martinet, Language 27: 333–42. Reprinted in Makkai (1972), 310–17.

Holowka, S. & L. A. Petitto (2002). “Left Hemisphere Cerebral Specialization for BabiesWhile Babbling.” Science 297: 1515.


322 References

Holowka, S., F. Brosseau-Lapre & L. A. Petitto (2002). “Semantic and ConceptualKnowledge Underlying Bilingual Babies’ First Signs and Words.” Language Learn-ing 52 (2): 205–62.

Hooper [Bybee], Joan (1976). An Introduction to Natural Generative Phonology. NewYork: Academic Press.

Hornstein, N. & David Lightfoot (eds.) (1981). Explanation in Linguistics: The LogicalProblem of Language Acquisition. London and New York: Longman.

Huang, C.-T. James (1981/82). “Move wh in a language without wh-movement.” TheLinguistic Review 1: 369–416.

(1982). “Logical Relations in Chinese and the Theory of Grammar.” Doctoral disser-tation, MIT, Cambridge, MA.

Hubel, D. (1978) “Vision and the Brain.” Bulletin of the American Academy of Arts andSciences 31 (7): 28.

Hubel, D. & T. Wiesel (1962). “Receptive Fields, Binocular Interaction and FunctionalArchitecture in the Cat’s Visual Cortex.” Journal of Physiology 160: 106–54.

Humboldt, Wilhelm von (1993). The Limits of State Action. Indianapolis: University ofIndiana Press.

(1999). On Language: On the Diversity of Human Language Construction and itsInfluence on the Mental Development of the Human Species, ed. Michael Losonsky,tr. Peter Heath. New York: Cambridge University Press.

Hume, D. (1978 (1739)). A Treatise on Human Nature. Oxford: Clarendon.Ingham, R. (1998). “Tense Without Agreement in Early Clause Structure.” Language

Acquisition 7: 51–81.Jackendoff, R. (1969). Some Rules of Semantic Interpretation for English. Doctoral

dissertation, MIT, Cambridge, MA.(1972). Semantic Interpretation in Generative Grammar. Cambridge, MA: MIT

Press.(1983). Semantics and Cognition. Cambridge, MA: MIT Press.

Jakobson, Roman & Morris Halle (1956). Fundamentals of Language. The Hague:Mouton.

Jakobson, Roman, C. Gunnar, M. Fant & M. Halle (1952). Preliminaries to SpeechAnalysis. MIT Acoustics Laboratory Technical Report No. 13. Reissued by MITPress, Cambridge, MA. 11th printing, 1976.

Jenkins, L. (2000). Biolinguistics: Exploring the Biology of Language. Cambridge:Cambridge University Press.

Jerne, N. K. (1967). “Antibodies and Learning: Selection Versus Instruction.” In G. C.Quarton, T. Melnechuk & F. O. Schmitt, eds. The Neurosciences: A Study Program.New York: Rockefeller University Press.

(1985). “The Generative Grammar of the Immune System” [Nobel lecture]. Science229: 1057–9.

Joanisse, M. & M. Seidenberg (1998). “Specific Language Impairment: A Deficit inGrammar or Processing?” Trends in Cognitive Sciences 2: 240–7.

Joos, Martin (ed.) (1957). Readings in Linguistics I. Chicago: University of ChicagoPress. [2nd edition New York: American Council of Learned Societies, 1958.]

Joshi, A. & B. Srinivas (1994). “Disambiguation of Super Parts-of-speech (or Supertags):Almost Parsing.” Proceedings of the 15th International Conference on Computa-tional Linguistics (COLING ‘94). Kyoto, Japan. 154–60.


References 323

Jusczyk, P. W. (1999). “Narrowing the Distance to Language: One Step at a Time.”Journal of Communication Disorders 32 (4): 207–22.

Kako, E. & L. R. Gleitman (in prep). “Information Sources for the Learning of Nouns.”Katz, Jerrold J. & Paul Postal (1964). An Integrated Theory of Linguistic Descriptions.

Cambridge, MA: MIT Press.Kay, Jonathan (2004). “Ethical Dilemmas.” Business (March 2004). Magazine published

by Canada’s National Post.Kaye, Jonathan D. (1974). “Opacity and Recoverability in Phonology.” Canadian

Journal of Linguistics 19: 134–49.Kaye, Jonathan D., Jean Lowenstamm & Jean-Roger Vergnaud (1985). “The Inter-

nal Structure of Phonological Elements: A Theory of Charm and Government.”Phonology Yearbook 2: 305–28.

Kean, Mary-Louise (1980). The Theory of Markedness in Generative Grammar. Ph.D.dissertation, MIT (1975). [Published Bloomington: Indiana University LinguisticsClub.]

Keenan, E. L. (1976). “Towards a Universal Definition of Subject.” In C. N. Li, ed.Subject and Topic. New York: Academic Press. 303–33.

Kenstowicz, Michael (1994). Phonology in Generative Grammar. Oxford: Blackwell.Kenstowicz, Michael & Charles Kisseberth (1977). Topics in Phonological Theory. New

York: Academic Press.Kiparsky, Paul (1968). “How Abstract is Phonology?” Bloomington: Indiana University

Linguistics Club. Reprinted in Kiparsky (1982b), 119–64.(1973). “Abstractness, Opacity, and Global Rules.” Bloomington: Indiana University

Linguistics Club. Also in O. Fujimura, ed. Three Dimensions of Linguistic Theory.Tokyo: TEC. 57–86.

(1982a). “From Cyclic Phonology to Lexical Phonology.” In Harry van der Hulst andNorval Smith, eds. The Structure of Phonological Representations (Part I). 131–76.Dordrecht: Foris.

(1982b). Explanation in Phonology. Dordrecht: Foris.(1985). “Some Consequences of Lexical Phonology.” Phonology Yearbook 2: 85–138.

Kosslyn, S. M., M. S. Gazzaniga, A. M. Galaburda & C. Rabin (1999). “HemisphericSpecialization.” In M. J. Zigmond, F. E. Bloom, S. C. Landis, J. L. Roberts & L. R.Squire, eds. Fundamental Neuroscience. San Diego: Academic Press.

Kovelman, I. & L. A. Petitto (2002). “Bilingual Babies’ Maturational and LinguisticMilestones as a Function of Their Age of First Exposure to Two Languages.”Published Abstracts of the 32nd Annual Meeting of the Society for Neuroscience.Orlando, FL.

(2003). “Bilingual Exposure at Different Ages in Childhood: Do They Exhibit ‘Stage-like’ Language Acquisition Similar to Young Monolinguals?” Published Abstractsof the International Symposium on Bilingualism. Tempe, AZ.

Kripke, Saul (1972). “Naming and Necessity.” In D. Davidson and G. Harman, eds.Semantics for Natural Language. Dordrecht: D. Reidel.

(1976). “A Puzzle About Belief.” In Avishai Margalit, ed. Meaning and Use.Dordrecht: D. Reidel.

Ladefoged, Peter & Ian Maddieson (1996). The Sounds of the World’s Languages.Oxford: Blackwell.

LaFeber, Walter (1967). The New Empire. Ithaca: Cornell University Press.


324 References

Landau, B. & L. R. Gleitman (1985). Language and Experience: Evidence From theBlind Child. Cambridge, MA: Harvard University Press.

Larson, Richard & Gabriel Segal (1995). Knowledge of Meaning: An Introduction toSemantic Theory. Cambridge, MA: MIT Press.

Lasnik, Howard (1972). “Analyses of Negation in English.” Doctoral dissertation, MIT,Cambridge, MA.

(1976). “Remarks on Coreference.” Linguistic Analysis 2: 1–22. [Reprinted in Lasnik,Essays on Anaphora. Dordrecht: Kluwer, 1989. 90–109.]

Lederer, A., H. Gleitman & L. R. Gleitman (1995). “Verbs of a Feather Flock Together:Semantic Information in the Structure of Maternal Speech.” In M. Tomasello &W. E. Merriman, eds. Beyond Names for Things: Young Children’s Acquisition ofVerbs. Hillsdale, NJ: Lawrence Erlbaum. 277–97.

Lee, J. & L. R. Naigles (2002). “Syntactic Bootstrapping with Missing Arguments:Mandarin Chinese.” Paper presented at the IXth International Congress for theAcquisition of Language, Madison, WI.

Lefebvre, C. (1998). Creole Genesis and the Acquisition of Grammar. Cambridge:Cambridge University Press.

Lely, H. K. J. van der (1996). “Specifically Language Impaired and Normally DevelopingChildren: Verbal Passive vs. Adjectival Passive Sentence Interpretation.” Lingua98: 243–72.

Levy, Y. & G. Kave (1999). “Language Breakdown and Linguistic Theory: A TutorialOverview.” Lingua 107: 95–143.

Lewis, Anthony (1987). “Freedom of the Press – Anthony Lewis Distinguishes BetweenBritain and America.” London Review of Books, Nov. 26.

Lewontin, R. (1990). “The Evolution of Cognition.” In D. N. Osherson and E. E. Smith,eds. An Invitation to Cognitive Science, vol. 3, Thinking. Cambridge, MA: MITPress. 229–46.

Li, P. (1994). “Subcategorization as a Predictor of Verb Meaning: Cross-language Studyin Mandarin.” Unpublished manuscript, University of Pennsylvania.

Liberman, A. M. & I. G. Mattingly (1989). “A Specialization for Speech Perception.”Science 243: 489–94.

Lidz, J., H. Gleitman & L. R. Gleitman (2003). “Understanding How Input Matters:Verb Learning and the Footprint of Universal Grammar.” Cognition 87: 151–78.

Lieberman, P. (2000). Human Language and Our Reptilian Brain: The Subcortical Basesof Speech, Syntax, and Thought. Cambridge, MA: Harvard University Press.

Lightfoot, D. W. (1999). The Development of Language: Acquisition, Change andEvolution. Oxford: Blackwell.

Linell, Per (1979). Psychological Reality in Phonology: A Theoretical Study. Cambridge:Cambridge University Press.

Locke, J. (1690). An Essay Concerning Human Understanding, ed. A. D. Woozley.Cleveland: Meridian Books, 1964.

Locke, J. L. (2000). “Movement Patterns in Spoken Language.” Science 288: 449–51.Ludlow, Peter (2002) “LF and Natural Logic: The Syntax of Directional Entailing Envi-

ronments.” In G. Preyer and G. Peter, eds. Logical Form and Language. Oxford:Oxford University Press.

Macedo, Donaldo (ed.) (2000). Chomsky on MisEducation. NewYork: Rowman andLittlefield.


References 325

MacNeilage, P. F. & B. Davis (2000). “On the Origin of Internal Structure of WordForms.” Science 288: 527–31.

Madison, James (n.d.). The Federalist Papers. [Available: federalistpapers.com]Makkai, Valerie Becker (ed.) (1972). Phonological Theory: Evolution and Current

Practice. New York: Holt, Rinehart and Winston.Manzini, M. R. & K. Wexler (1987). “Parameters, Binding Theory and Learnability.”

Linguistic Inquiry 18: 413–44.Marantz, Alec (1997). “No Escape from Syntax: Don’t Try Morphological Analysis

in the Privacy of Your Own Lexicon.” Penn Working Papers in Linguistics 4:2.Philadelphia: University of Pennsylvania.

Marcus, G. F., S. Vijayan, S. Bandi & P. M. Vishton (1999). “Rule Learning by Seven-month-old Infants.” Science 283: 77–80.

Marlett, Stephen A. (1981). “The Abstract Consonant in Seri.” In Proceedings of theBerkeley Linguistics Society 7. Berkeley: University of California at Berkeley.154–65.

Marlett, Stephen A., & Joseph P. Stemberger (1983). “Empty Consonants in Seri.”Linguistic Inquiry 14: 617–39.

May, Robert (1977). “The Grammar of Quantification.” Doctoral dissertation, MIT,Cambridge, MA.

McCawley, James D. (1968). “The Role of Semantics in Grammar.” In Emmon Bach &Robert T. Harms, eds. Universals in Linguistic Theory. New York: Holt, Rinehartand Winston. 124–69.

McGilvray, James (1998). “Meanings are Syntactically Individuated and Found in theHead.” Mind and Language 13 (2): 225–80.

(1999). Chomsky. Cambridge: Polity.(2002a). “MOPs: the Science of Concepts.” In W. Hinzen and H. Rott, eds. Belief and

Meaning, Essays at the Interface. Frankfurt am Main: Hansel-Hohenhausen.(2002b). “Introduction for Cybereditions.” In Chomsky, Cartesian Linguistics, 2nd

edition. Christchurch, NZ: Cybereditions.McGinn, Colin (1994a). “The Problem of Philosophy. Philosophical Studies 76: 133–56.

(1994b). “Reply to Carol Rovane.” Philosophical Studies 76: 169–74.Miringoff, Marc & Marque-Luisa Miringoff (1999). The Social Health of the Nation.

New York: Oxford University Press.Monk, Ray (1997). Bertrand Russell. The Spirit of Solitude. London: Vintage.Monod, J. (1972). Chance and Necessity. London: Collins.Moravcsik, Julius (1975). “Aitia as Generative Factor in Aristotle’s Philosophy of

Language.” Dialogue 14: 622–36.Naigles, L. (1990). “Children Use Syntax to Learn Verb Meanings.” Journal of Child

Language 17: 357–74.Naigles, L. G. & E. T. Kako (1993). “First Contact in Verb Acquisition: Defining a Role

for Syntax.” Child Development 64: 1665–87.Naigles, L. R. & P. Terrazas (1998). “Motion-verb Generalizations in English

and Spanish: Influences of Language and Syntax.” Psychological Science 9:363–9.

Naigles, L., H. Gleitman & L. R. Gleitman (1992). “Children Acquire Word MeaningComponents from Syntactic Evidence.” In E. Dromi, ed. Language and Cognition:A Developmental Perspective. Norwood, NJ: Ablex. 104–44.


326 References

Navarro, Vincente (1994). The Politics of Health Policy: The U.S. Reforms, 1980–1994.Cambridge, MA: Blackwell.

Newport, E. L. (1990). “Maturational Constraints on Language Learning.” CognitiveScience 14: 11–28.

(1999). “Reduced Input in the Acquisition of Signed Languages: Contributions to theStudy of Creolization.” In DeGraff (1999).

Norton, E. S., S. A. Baker & L. A. Petitto (2003). “Bilingual Infants’ Perception ofHandshapes in American Sign Language.” Poster presented at the 6th AnnualSummer Undergraduate Workshop, Institute for Research in Cognitive Science,University of Pennsylvania, Philadelphia, PA.

Oakley, Allan (1994). Classical Economic Man: Human Agency and Methodology inthe Political Economy of Adam Smith and J. S. Mill. Vermont: Edward Elgar.

Outram, Dorinda (1995). The Enlightenment. Cambridge: Cambridge UniversityPress.

Papafragou, A., C. Massey, & L. R. Gleitman (2002). “Shake, Rattle, ’n’ Roll: TheRepresentation of Motion in Language and Cognition.” Cognition 84: 189–219.

Penhune, V. B., R. Cismaru, R. Dorsaint-Pierre, L. A. Petitto & R. J. Zatorre (2003).“The Morphometry of Auditory Cortex in the Congenitally Deaf Measured UsingMRI.” NeuroImage 20 (2): 1215–25.

Petitto, L. A. (2000). “On the Biological Foundations of Human Language.” In K.Emmorey & H. Lane, eds. The Signs of Language Revisited: An Anthology inHonor of Ursula Bellugi and Edward Klima. Mahwah, NJ: Lawrence Erlbaum.447–71.

Petitto, L. A. & S. Holowka (2002). “Does Early Simultaneous Bilingual LanguageExposure Cause Children to be Language Delayed and Confused: Special Insightsfrom Bilingual Babies and Young Children Acquiring a Signed and a SpokenLanguage.” Sign Language Studies 3 (1): 4–33.

Petitto, L. A. & I. Kovelman (2003). “The Bilingual Paradox: How Signing-speakingBilingual Children Help Us Resolve Bilingual Issues and Teach us About the Brain’sMechanisms Underlying All Language Acquisition.” Learning Languages 8 (3):5–18.

Petitto, L. A. & P. Marentette (1991). “Babbling in the Manual Mode: Evidence for theOntogeny of Language.” Science 251: 1493–6.

Petitto, L. A., R. J. Zatorre, K. Gauna, E. J. Nikelski, D. Dostie & A. C. Evans (2000).“Speech-like Cerebral Activity in Profoundly Deaf People While Processing SignedLanguages: Implications for the Neural Basis of all Human Language.” Proceedingsof the National Academy of Sciences 97 (25): 13961–6.

Petitto, L. A., M. Katerelos, B. Levy, K. Gauna, K. Tetreault & V. Ferraro (2001).“Bilingual Signed and Spoken Language Acquisition From Birth: Implications forthe Mechanisms Underlying Early Bilingual Language Acquisition.” Journal ofChild Language 28 (2): 1–44.

Petitto, L. A., S. Holowka, L. Sergio & D. Ostry (2002). “Language Rhythms in Babies’Hand Movements.” Nature 413: 35–6.

Petitto, L. A., I. Kovelman & U. Harasymowycz (2003). “Bilingual LanguageDevelopment: Does Learning the New Damage the Old?” Published Abstractsof the Society for Research in Child Development Biennial Meeting. Tampa, FL.

Petitto, L. A., S. Holowka, L. Sergio, D. Ostry & B. Levy (2004). “Baby Hands thatMove to the Rhythm of Language: Hearing Babies Acquiring Sign LanguagesBabble Silently on the Hands.” Cognition 93: 43–73.


References 327

Piaget, J. & B. Inhelder (1968) The Psychology of the Child. London: Routledge.Piattelli-Palmarini, M. (ed.) (1980). Language and Learning: The Debate Between Jean

Piaget and Noam Chomsky. London: Routledge and Kegan Paul.(1986) “The Rise of Selective Theories: A Case Study and Some Lessons from

Immunology.” In W. Demopoulos & A. Marras, eds. Language Learning andConcept Acquisition: Foundational Issues. Norwood, NJ: Ablex.

(1989). “Evolution, Selection and Cognition: From Learning to Parameter Setting inBiology and in the Study of Language.” Cognition 31: 1–44.

Pietroski, Paul (2002). “Meaning Before Truth.” In G. Preyer and G. Peter, eds.Contextualism in Philosophy. Oxford: Oxford University Press.

(2003). “The Character of Natural Language Semantics.” In A. Barber, ed. Episte-mology of Language. Oxford: Oxford University Press.

(2005). Events and Semantic Architecture. Oxford: Oxford University Press.Pilger, John (1998). Hidden Agendas. London: Vintage.Pinker, S. (1984). Language Learnability and Language Development. Cambridge, MA:

Harvard University Press.(1989). Learnability and Cognition. Cambridge, MA: MIT Press.(1993). “The Central Problem for the Psycholinguist.” In G. Harman, ed. Conceptions

of the Human Mind: Papers in Honor of George Miller. Hillsdale, NJ: Erlbaum.(1994). The Language Instinct. New York: William Morrow.(1999). “Out of the Minds of Babies.” Science 283: 40–1.(2002). The Blank Slate. New York: Penguin Putnum.

Pinker, S. & P. Bloom (1990). “Natural Language and Natural Selection.” Behavioraland Brain Sciences 13 (4): 707–84.

Popper, K. (1963). Conjectures and Refutations: The Growth of Scientific Knowledge.London: Routledge and Kegan Paul.

Postal, Paul M. (1968). Aspects of Phonological Theory. New York: Harper & Row.(1972). “The Best Theory.” In P. S. Peters, ed. Goals of Linguistic Theory. Englewood

Cliffs, NJ: Prentice-Hall. 131–70.Premack, D. & G. Woodruff (1978). “Chimpanzee Problem-solving: A Test for

Comprehension.” Science 202: 532–35.Premack, D. (1980). “Representational Capacity and Accessibility of Knowledge: The

Case of Chimpanzees.” In Piattelli-Palmarini (1980).Premack, D. (1990). “Words: What Are They, and Do Animals Have Them?” Cognition

37 (3): 197–212.Pullum, G. K. (1996). “Learnability, Hyperlearning, and the Poverty of the Stimulus.”

In J. Johnson, M. L. Juge & J. L. Moxley, eds. Proceedings of the 22nd AnnualMeeting: General Session and Parasession on the Role of Learnability in Gram-matical Theory. Berkeley: Berkeley Linguistics Society. 498–513.

Pustejovsky, James (1995). The Generative Lexicon. Cambridge, MA: MIT Press.Putnam, Hilary (1967). “The Innateness Hypothesis and Explanatory Models in

Linguistics.” Synthese 17: 12–22. Reprinted in Putnam, Mind, Language andReality. Cambridge: Cambridge University Press, 1975.

(1975). “The Meaning of ‘Meaning’.” In K. Gunderson, ed. Language, Mind andKnowledge: Minnesota Studies in the Philosophy of Science, 7. Minneapolis:University of Minnesota Press.

Quine, W. (1960). Word and Object. New York: Wiley.(1972). “Methodological Reflections on Current Linguistic Theory.” In D. Davidson

& G. Harman, eds. Semantics of Natural Language. Dordrecht: Reidel. 442–54.


328 References

Rai, Milan (1995). Chomsky’s Politics. London: Verso.Rappaport-Hovav, M. & B. Levin (1988). “What to Do With Theta-roles.” In W. Wilkins,

ed. Syntax and Semantics, vol. 21: Thematic Relations. New York: Academic. 7–36.Raskin, Markus G. & Herbert J. Bernstein (eds.) (1987). New Ways of Knowing:

The Sciences, Society, and Reconstructive Knowledge. Towota, NJ: Rowman andLittlefield.

Rice, Keren (2002). “Vowel Place Contrasts.” In Mengistu Amberber & Peter Collins,eds. Language Universals and Variation. Westport, CT, London: Praeger. 239–70.

Rizzolatti, G., V. Gallese, L. Fadiga & L. Fogassi (1996). “Action Recognition in thePremotor Cortex.” Brain 119: 592–609.

Roca, Iggy & Wyn Johnson (1999). A Course in Phonology. Oxford: Blackwell.Ross, John Robert (1967). “Constraints on Variables in Syntax.” Doctoral dissertation,

MIT, Cambridge, MA. Published as Infinite Syntax! Norwood, NJ: Ablex, 1986.Russell, Bertrand (1920). The Practice and Theory of Bolshevism. London: Allen and

Unwin.Sapir, Edward (1933). “La realite psychologique des phonemes.” Journal de Psycholo-

gie Normale et Pathologique 30: 247–65. Reprinted as “The Psychological Real-ity of Phonemes.” In D. Mandelbaum, ed. Selected Writings of Edward Sapirin Language, Culture, and Personality. Berkeley and Los Angeles: University ofCalifornia Press (1949), 46–60. Also in Makkai (1972), 22–31.

Seidenberg, M. S. & L. A. Petitto (1979). “Signing Behavior in Apes: A Critical Review.”Cognition 7: 177–215.

(1987). “Communication, Symbolic Communication, and Language in Child andChimpanzee: Comment on Savage-Rumbaugh, McDonald, Sevcik, Hopkins, andRupert (1986).” Journal of Experimental Psychology 116 (3): 279–87.

Sellars, Wilfrid (1974). “Meaning as Functional Classification.” Synthese 27: 417–37.

Shipley, E. F., I. F. Kuhn & E. C. Madden (1983). “Mothers’ Use of SuperordinateCategory Terms.” Journal of Child Language 10: 571–88.

Sieratzki, J. & B. Woll (2002). “Toddling into Language: Precocious Language Devel-opment in Motor-impaired Children with Spinal Muscular Atrophy.” Lingua 112(6): 423–33.

Skinner, B. F. (1938). Behavior of Organisms. New York: Appleton-Century-Crofts.(1957). Verbal Behavior. New York: Appleton-Century-Crofts.

Slobin, D. I. (1982). “Universal and Particular in the Acquisition of Language.” InE. Wanner & L. R. Gleitman, eds. Language Acquisition: The State of the Art.New York: Cambridge University Press. 128–70.

Slobin, D. I. (2001). “Form-function Relations: How do Children Find Out What TheyAre?” In M. Bowerman & S. C. Levinson, eds. Language Acquisition and Concep-tual Development. New York: Cambridge University Press. 406–49.

Smith, Adam (1759). The Theory of Moral Sentiments. Available athttp://www.adamsmith.org/smith/tms/tms-index.htm

(1990). An Inquiry into the Wealth of Nations. Chicago: Encyclopaedia Britannica.Smith, N. (1998). “Jackdaws, Sex and Language Acquisition.” Glot International

3 (7): 7.(1999). Chomsky: Ideas and Ideals. Cambridge: Cambridge University Press.(2000). “Foreword” to Chomsky (2000a). vi–xvi.


References 329

(2003) “Dissociation and Modularity: Reflections on Language and Mind.” InM. Banich & M. Mack, eds. Mind, Brain and Language. Hillsdale, NJ: LawrenceErlbaum. 87–111.

Smith, N. & I.-M. Tsimpli (1995). The Mind of a Savant: Language-learning andModularity. Oxford: Blackwell.

Snedeker, J. & L. R. Gleitman (in press). “Why It Is Hard to Label Our Concepts.” Toappear in D. G. Hall & S. R. Waxman, eds. Weaving a Lexicon. Cambridge, MA:MIT Press.

Snedeker, J., K. Thorpe & J. Trueswell (2001). “On Choosing the Parse With the Scene:The Role of Visual Context and Verb Bias in Ambiguity Resolution.” Proceedings ofthe 22nd Annual Conference of the Cognitive Science Society, Edinburgh, Scotland.

Sperry, R. (1968). “Plasticity of Neural Maturation.” Developmental Biology Supplement2: 306–27.

Strange, Susan (1998). Mad Money: When Markets Outgrow Governments. Ann Arbor:University of Michigan Press.

Supalla, S. (1990). “Segmentation of Manually Coded English: Problems in the Mappingof English in the Visual/gestural Mode.” Ph.D. dissertation, University of Illinois.

Swadesh, Morris & Charles F. Voegelin (1939). “A Problem in Phonological Alterna-tion.” Language 15: 1–10. Reprinted in Joos (1957), 88–92.

Talbot, Margaret (2001). “The Lives They Lived.” The New York Times. Final edition,Dec. 30, section 6.

Talmy, L. (1985). “Lexicalization Patterns: Semantic Structure in Lexical Forms.” InT. Shopen, ed. Language Typology and Syntactic Description. New York:Cambridge University Press. 57–149.

Terrace, H. S., L. A. Petitto, R. J. Sanders & T. G. Bever (1979). “Can an Ape Create aSentence?” Science 206: 891–902.

Tomasello, M. (2000). “Do Young Children Have Adult Syntactic Competence?”Cognition 74: 209–253.

Thompson, W. D’Arcy (1917). On Growth and Form. Cambridge: Cambridge UniversityPress.

Tomblin, J. B. (1997). “Epidemiology of Specific Language Impairment.” In M. Gopnik,ed. The Inheritance and Innateness of Grammars. New York: Oxford UniversityPress.

Tranel, Bernard (1981). “Concreteness in Generative Phonology: Evidence fromFrench.” Berkeley, CA: University of California Press.

Trueswell, J. C. & A. E. Kim (1998). “How to Prune a Garden Path by Nipping It in theBud: Fast Priming of Verb Argument Structure.” Journal of Memory & Language39: 102–23.

Turing, A. M. (1950). “Computing Machinery and Intelligence.” Mind 59: 433–560.(1952). “The Chemical Basis of Morphogenesis.” Philosophical Transactions of the

Royal Society of London. 37–72.Twaddell, F. (1935). “On Defining the Phoneme.” Language Monograph, no. 16.

Reprinted in Joos (1957), 55–80.Ullmann, Stephen (1957). The Principles of Semantics, 2nd edition. Oxford: Basil

Blackwell.Uriagereka, J. (1998). Rhyme and Reason: An Introduction to Minimalist Syntax.

Cambridge, MA: MIT Press.


330 References

(1999). “Multiple Spell-out.” In Samuel David Epstein & Norbert Hornstein, eds.Working Minimalism. Cambridge, MA: MIT Press. 251–82.

Viner, Jacob (1991). Essays on the Intellectual History of Economics, ed. DouglasAlwrin. Princeton: Princeton University Press.

Waxman, S. R. & D. B. Markow (1995). “Words as Invitations to Form Categories:Evidence From 12- to 13-month-old Infants.” Cognitive Psychology 29: 257–302.

Wernicke, C. (1874). Der aphasische Symptomencomplex. Breslau: Kohn & Weigert.West, G. B., J. H. Brown & B. J. Enquist (2000). “The Origin of Universal Scaling Laws

in Biology”. In J. H. Brown & G. B. West, eds. Scaling in Biology. Oxford: OxfordUniversity Press. 87–112.

Wiggins, David (1997). “Languages as Social Objects”. Philosophy 72: 499–524.Wilkins, Peter (1997). Noam Chomsky: On Power, Knowledge and Human Nature.

Basingstoke: Macmillan.Wilson, James G. (1996). “Commentary: Noam Chomsky and Judicial Review.” Cleve-

land State Law Review 439.(2002). The Imperial Republic: A Structural History of American Constitutionalism

From the Colonial Era to the Beginning of the Twentieth Century. Aldershot, UK:Ashgate Publishing.

Wittgenstein, Ludwig (1963). Philosophical Investigations, trans. G. E. M. Anscombe.Oxford: Basil Blackwell.

Woodward, A. L. & E. M. Markman (1998). “Early Word learning.” In W. Damon (seriesed.); D. Kuhn & R. S. Siegler (vol. eds.) Handbook of Child Psychology, vol. 2:Cognition, Perception, and Language, 5th edition. New York: Wiley. 371–420.

Zatorre, R. J. & J. R. Binder (2000). “Functional Imaging of the Chemical Senses.”In A. Toga & J. C Mazziota, eds. Brain Mapping: The Systems. San Diego, CA:Academic Press. 124–50.


The Cambridge Companion to Chomsky

Documents